Bubble sheet multiple choice scanner and test grader using OMR, Python and OpenCV

小灰灰 2021-10-23 11:26 215阅读 0赞

转自国外opencv教育网站免费教程：

https://www.pyimagesearch.com/2016/10/03/bubble-sheet-multiple-choice-scanner-and-test-grader-using-omr-python-and-opencv/

# 使用OMR，Python和OpenCV的气泡板多选扫描仪和测试分级机 #

作者： [Adrian Rosebrock][] 于 2016 年 10月3日 在 [图像处理][Link 1]，[教程中][Link 2]

![图14：使用计算机视觉识别气泡板检查。][14]

在过去的几个月里，我已经收到大量的请求登陆我的收件箱，以使用计算机视觉和图像处理技术构建一个泡泡板/类似Scantron的测试阅读器。

虽然我在机器学习和深度学习这个系列中玩得很开心，但如果我说这个小小的项目并不是一个短暂的，欢迎的休息时间，我会  *说谎*。我最喜欢运行PyImageSearch博客的部分之一是演示如何使用计算机视觉构建  问题的*实际*解决方案。

事实上，这个项目  *如此特别的*原因在于我们将结合以前*许多*博客文章中的技术  ，包括[构建文档扫描程序][Link 3]，[轮廓排序][Link 4]和[透视变换][Link 5]。利用从以前的帖子中获得的知识，我们将能够快速完成这个气泡板扫描仪和测试分级机。

你看，上周五下午我快速上了一个示例泡泡试卷，打印出几份副本，  *然后开始编写实际实现的编码。*

总的来说，我对这个实现非常满意，我认为你绝对可以使用这个bubble bubble grader / OMR系统作为你自己项目的起点。

**要了解有关利用计算机视觉，图像处理和OpenCV自动评定气泡测试表的更多信息，  *请继续阅读。***

寻找这篇文章的源代码？  
跳到下载部分。

## 使用OMR，Python和OpenCV的气泡板扫描仪和测试分级机 ##

在本博文的其余部分，我将讨论*光学标记识别*（OMR）究竟  是什么。然后，我将演示如何使用*严格的*计算机视觉和图像处理技术以及OpenCV库来实现气泡板测试扫描仪和分级机  。

一旦我们实施了OMR系统，我将在一些示例考试中提供我们的考试评分员的样本结果，包括那些充满邪恶意图的考试。

最后，我将讨论当前气泡板扫描仪系统的一些缺点，以及我们如何在未来的迭代中对其进行改进。

### 什么是光学标记识别（OMR）？ ###

光学标记识别（简称OMR）是  *自动*分析人工标记文档并解释其结果的过程。

可以说，OMR最着名，易于识别的形式是  ***泡泡纸多项选择测试***，与你在小学，初中甚至高中时所采用的测试不同。

如果您不熟悉“气泡纸测试”或“Scantron测试”的商标/公司名称，它们只是您作为学生的多项选择测试。考试中的每个问题都是多项选择 - 您使用＃2铅笔标记对应于正确答案的“气泡”。

在填写大学录取申请之前，您经历过的最值得注意的泡泡板测试（至少在美国）是在高中期间参加SAT考试。

我  *相信* SAT使用Scantron提供的软件来执行OMR和年级学生考试，但我很容易在那里出错。我只注意到这一点，因为Scantron在美国98％以上的学区使用。

简而言之，我想说的是，光学标记识别有一个  *巨大的市场*，并且能够对人类标记的形式和考试进行评级和解释。

### 使用OMR，Python和OpenCV实现气泡板扫描仪和分级器 ###

现在我们已经了解了OMR的基础知识，让我们使用Python和OpenCV构建一个计算机视觉系统，可以  *读取*和  *评估*气泡板测试。

当然，我将提供大量的视觉示例图像，以便您可以  *准确理解我正在应用的技术*以及  *我使用它们的原因。*

下面我列出了一个填写泡泡板考试的例子，我已经为这个项目整理了一个：

![图1：我们将在开发测试扫描仪软件时使用的填充气泡板的示例。][1]

**图1：**我们将在开发测试扫描仪软件时使用的填充气泡板的示例。

当我们完成构建测试分级器的步骤时，我们将使用它作为示例图像。在本课程的后面部分，您还可以找到其他样本考试。

我还将*空白考试模板*作为.PSD（Photoshop）文件包含在内，  以便您可以根据需要进行修改。您可以使用本文底部的  ***“下载”***部分下载代码，示例图像和模板文件。

#### 构建气泡板扫描仪和平地机的7个步骤 ####

这篇博客文章的目标是使用Python和OpenCV构建一个气泡表扫描程序和测试分级程序。

为实现这一目标，我们的实施需要满足以下7个步骤：

*  **步骤1：**检测图像中的检查。
 *  **步骤＃2：**应用透视变换以提取自上而下的鸟瞰视图。
 *  **步骤＃3：**从透视变换检查中提取气泡集（即可能的答案选项）。
 *  **步骤＃4：**将问题/气泡排序成行。
 *  **步骤＃5：**确定每行的标记（即“冒泡”）答案。
 *  **步骤＃6：**在我们的答案键中查找正确的答案，以确定用户的选择是否正确。
 *  **第7步：**重复考试中的所有问题。

本教程的下一部分将介绍我们算法的实际  *实现*。

#### 使用Python和OpenCV实现泡泡板扫描程序 ####

要开始，请打开一个新文件，将其命名为 test\_grader 。py  ，让我们开始工作：

<div id="crayon-5d52be9f4eb30930014621" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

<table> 
 <tbody> 
  <tr> 
   <td><br> 
    <div> 
     <div>
       1 
     </div> 
     <div>
       2 
     </div> 
     <div>
       3 
     </div> 
     <div>
       4 
     </div> 
     <div>
       5 
     </div> 
     <div>
       6 
     </div> 
     <div>
       7 
     </div> 
     <div>
       8 
     </div> 
     <div>
       9 
     </div> 
     <div>
       10 
     </div> 
     <div>
       11 
     </div> 
     <div>
       12 
     </div> 
     <div>
       13 
     </div> 
     <div>
       14 
     </div> 
     <div>
       15 
     </div> 
     <div>
       16 
     </div> 
     <div>
       17 
     </div> 
    </div><br> </td> 
   <td> 
    <div> 
     <div> 
      <span># import the necessary packages</span> 
     </div> 
     <div> 
      <span>from</span> 
      <span> </span> 
      <span>imutils</span> 
      <span>.</span> 
      <span>perspective </span> 
      <span>import</span> 
      <span> </span> 
      <span>four_point_transform</span> 
     </div> 
     <div> 
      <span>from</span> 
      <span> </span> 
      <span>imutils </span> 
      <span>import</span> 
      <span> </span> 
      <span>contours</span> 
     </div> 
     <div> 
      <span>import</span> 
      <span> </span> 
      <span>numpy </span> 
      <span>as</span> 
      <span> </span> 
      <span>np</span> 
     </div> 
     <div> 
      <span>import</span> 
      <span> </span> 
      <span>argparse</span> 
     </div> 
     <div> 
      <span>import</span> 
      <span> </span> 
      <span>imutils</span> 
     </div> 
     <div> 
      <span>import</span> 
      <span> </span> 
      <span>cv2</span> 
     </div> 
     <div>
       &nbsp; 
     </div> 
     <div> 
      <span># construct the argument parse and parse the arguments</span> 
     </div> 
     <div> 
      <span>ap</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>argparse</span> 
      <span>.</span> 
      <span>ArgumentParser</span> 
      <span>(</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>ap</span> 
      <span>.</span> 
      <span>add_argument</span> 
      <span>(</span> 
      <span>"-i"</span> 
      <span>,</span> 
      <span> </span> 
      <span>“–image”</span> 
      <span>,</span> 
      <span> </span> 
      <span>required</span> 
      <span>=</span> 
      <span>True</span> 
      <span>,</span> 
     </div> 
     <div> 
      <span> </span> 
      <span>help</span> 
      <span>=</span> 
      <span>“path to the input image”</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>args</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>vars</span> 
      <span>(</span> 
      <span>ap</span> 
      <span>.</span> 
      <span>parse_args</span> 
      <span>(</span> 
      <span>)</span> 
      <span>)</span> 
     </div> 
     <div>
       &nbsp; 
     </div> 
     <div> 
      <span># define the answer key which maps the question number</span> 
     </div> 
     <div> 
      <span># to the correct answer</span> 
     </div> 
     <div> 
      <span>ANSWER_KEY</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>{ </span> 
      <span>0</span> 
      <span>:</span> 
      <span> </span> 
      <span>1</span> 
      <span>,</span> 
      <span> </span> 
      <span>1</span> 
      <span>:</span> 
      <span> </span> 
      <span>4</span> 
      <span>,</span> 
      <span> </span> 
      <span>2</span> 
      <span>:</span> 
      <span> </span> 
      <span>0</span> 
      <span>,</span> 
      <span> </span> 
      <span>3</span> 
      <span>:</span> 
      <span> </span> 
      <span>3</span> 
      <span>,</span> 
      <span> </span> 
      <span>4</span> 
      <span>:</span> 
      <span> </span> 
      <span>1</span> 
      <span>}</span> 
     </div> 
    </div></td> 
  </tr> 
 </tbody> 
</table>

在  **第2-7行，**我们导入了我们所需的Python包。

您应该已经在系统上安装了OpenCV和Numpy，但是您  *可能*没有最新版本的[imutils][]，我的便捷功能集使您可以更轻松地执行基本的图像处理操作。要安装 imutils   （或升级到最新版本），只需执行以下命令：

<div id="crayon-5d52be9f4eb39832383291" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Shell</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;">$ pip install --upgrade imutils</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num" data-line="crayon-5d52be9f4eb39832383291-1">1</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line" id="crayon-5d52be9f4eb39832383291-1"><span class="crayon-sy">$</span><span class="crayon-h"> </span><span class="crayon-e">pip </span><span class="crayon-v">install</span><span class="crayon-h"> </span><span class="crayon-o">--</span><span class="crayon-e">upgrade </span><span class="crayon-v">imutils</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

**第10-12行**解析了我们的命令行参数。我们这里只需要一个开关， \- image  ，这是输入气泡板测试图像的路径，我们将对其进行分级以确保正确性。

**第17行**然后定义我们的 ANSWER\_KEY  。

正如变量名称所示， ANSWER\_KEY   提供*问题编号的*整数映射  到*正确气泡*的  *索引。*

在这种情况下，键  *0*表示第  *一个问题*，而值  *1*表示*“B”*表示  正确答案（因为  *“B”*是 字符串  *“ABCDE”中*的索引  *1*）。作为第二个例子，考虑的关键  *1*映射到值  *4* \-这将表明，回答第二个问题是  *“E”*。

为方便起见，我在这里用简明的英文写了整个答案的关键：

*  **问题＃1：** B
 *  **问题＃2：**  E
 *  **问题3：**  A
 *  **问题＃4：**  D
 *  **问题＃5：**  B

接下来，让我们预处理输入图像：

<div id="crayon-5d52be9f4eb49534355560" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

<table> 
 <tbody> 
  <tr> 
   <td><br> 
    <div> 
     <div>
       19 
     </div> 
     <div>
       20 
     </div> 
     <div>
       21 
     </div> 
     <div>
       22 
     </div> 
     <div>
       23 
     </div> 
     <div>
       24 
     </div> 
    </div><br> </td> 
   <td> 
    <div> 
     <div> 
      <span># load the image, convert it to grayscale, blur it</span> 
     </div> 
     <div> 
      <span># slightly, then find edges</span> 
     </div> 
     <div> 
      <span>image</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>imread</span> 
      <span>(</span> 
      <span>args</span> 
      <span>[</span> 
      <span>“image”</span> 
      <span>]</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>gray</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>cvtColor</span> 
      <span>(</span> 
      <span>image</span> 
      <span>,</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>COLOR_BGR2GRAY</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>blurred</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>GaussianBlur</span> 
      <span>(</span> 
      <span>gray</span> 
      <span>,</span> 
      <span> </span> 
      <span>(</span> 
      <span>5</span> 
      <span>,</span> 
      <span> </span> 
      <span>5</span> 
      <span>)</span> 
      <span>,</span> 
      <span> </span> 
      <span>0</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>edged</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>Canny</span> 
      <span>(</span> 
      <span>blurred</span> 
      <span>,</span> 
      <span> </span> 
      <span>75</span> 
      <span>,</span> 
      <span> </span> 
      <span>200</span> 
      <span>)</span> 
     </div> 
    </div></td> 
  </tr> 
 </tbody> 
</table>

在  **第21行， **我们从磁盘加载图像，然后将其转换为灰度（**第22行**），并模糊它以减少高频噪声（**第23行**）。

然后我们在**第24行**应用Canny边缘检测器  来查找检查的  *边缘/轮廓*。

下面我在应用边缘检测后包含了我们考试的屏幕截图：

![图2：在我们的考试中应用边缘检测整齐地揭示了论文的大纲。][2]

**图2：**在我们的考试中应用边缘检测整齐地揭示了论文的大纲。

注意如何将文档的边缘被  *明确定义*，与  *考试的所有四个顶点*存在在图像中。

在我们的下一步中获取文档的这个轮廓是  *非常重要的*，因为我们将它用作标记来将透视变换应用于考试，从而获得文档的自上而下的鸟瞰视图：

<div id="crayon-5d52be9f4eb4e104799191" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# loop over the sorted contours
    for c in cnts:
    	# approximate the contour
    	peri = cv2.arcLength(c, True)
    	approx = cv2.approxPolyDP(c, 0.02 * peri, True)
    
    	# if our approximated contour has four points,
    	# then we can assume we have found the paper
    	if len(approx) == 4:
    		docCnt = approx
    		break</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-26">26</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-27">27</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-28">28</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-29">29</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-30">30</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-31">31</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-32">32</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-33">33</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-34">34</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-35">35</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-36">36</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-37">37</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-38">38</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-39">39</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-40">40</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-41">41</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-42">42</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-43">43</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-44">44</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-45">45</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-46">46</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-47">47</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb4e104799191-48">48</div><div class="crayon-num" data-line="crayon-5d52be9f4eb4e104799191-49">49</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-26"><span class="crayon-c"># find contours in the edge map, then initialize</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-27"><span class="crayon-c"># the contour that corresponds to the document</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-28"><span class="crayon-v">cnts</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">findContours</span><span class="crayon-sy">(</span><span class="crayon-v">edged</span><span class="crayon-sy">.</span><span class="crayon-k ">copy</span><span class="crayon-sy">(</span><span class="crayon-sy">)</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-v">RETR_EXTERNAL</span><span class="crayon-sy">,</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-29"><span class="crayon-h">	</span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-v">CHAIN_APPROX_SIMPLE</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-30"><span class="crayon-v">cnts</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">imutils</span><span class="crayon-sy">.</span><span class="crayon-e">grab_contours</span><span class="crayon-sy">(</span><span class="crayon-v">cnts</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-31"><span class="crayon-v">docCnt</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-t">None</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-32"> </div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-33"><span class="crayon-c"># ensure that at least one contour was found</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-34"><span class="crayon-st">if</span><span class="crayon-h"> </span><span class="crayon-k ">len</span><span class="crayon-sy">(</span><span class="crayon-v">cnts</span><span class="crayon-sy">)</span><span class="crayon-h"> </span><span class="crayon-o">></span><span class="crayon-h"> </span><span class="crayon-cn">0</span><span class="crayon-o">:</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-35"><span class="crayon-h">	</span><span class="crayon-c"># sort the contours according to their size in</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-36"><span class="crayon-h">	</span><span class="crayon-c"># descending order</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-37"><span class="crayon-h">	</span><span class="crayon-v">cnts</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-k ">sorted</span><span class="crayon-sy">(</span><span class="crayon-v">cnts</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">key</span><span class="crayon-o">=</span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-v">contourArea</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">reverse</span><span class="crayon-o">=</span><span class="crayon-t">True</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-38"> </div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-39"><span class="crayon-h">	</span><span class="crayon-c"># loop over the sorted contours</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-40"><span class="crayon-h">	</span><span class="crayon-st">for</span><span class="crayon-h"> </span><span class="crayon-i">c</span><span class="crayon-h"> </span><span class="crayon-st">in</span><span class="crayon-h"> </span><span class="crayon-v">cnts</span><span class="crayon-o">:</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-41"><span class="crayon-h">		</span><span class="crayon-c"># approximate the contour</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-42"><span class="crayon-h">		</span><span class="crayon-v">peri</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">arcLength</span><span class="crayon-sy">(</span><span class="crayon-v">c</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-t">True</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-43"><span class="crayon-h">		</span><span class="crayon-v">approx</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">approxPolyDP</span><span class="crayon-sy">(</span><span class="crayon-v">c</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">0.02</span><span class="crayon-h"> </span><span class="crayon-o">*</span><span class="crayon-h"> </span><span class="crayon-v">peri</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-t">True</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-44"> </div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-45"><span class="crayon-h">		</span><span class="crayon-c"># if our approximated contour has four points,</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-46"><span class="crayon-h">		</span><span class="crayon-c"># then we can assume we have found the paper</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-47"><span class="crayon-h">		</span><span class="crayon-st">if</span><span class="crayon-h"> </span><span class="crayon-k ">len</span><span class="crayon-sy">(</span><span class="crayon-v">approx</span><span class="crayon-sy">)</span><span class="crayon-h"> </span><span class="crayon-o">==</span><span class="crayon-h"> </span><span class="crayon-cn">4</span><span class="crayon-o">:</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb4e104799191-48"><span class="crayon-h">			</span><span class="crayon-v">docCnt</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-e">approx</span></div><div class="crayon-line" id="crayon-5d52be9f4eb4e104799191-49"><span class="crayon-e">			</span><span class="crayon-st">break</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

现在我们有了考试大纲，我们应用了 cv2 。findContours   函数用于查找与考试本身对应的行。

我们通过在**37号线上**按照它们的  *面积*（从大到小）对  轮廓进行分类来确定这一点（当然，确保在**34号线上**找到至少一个轮廓  ）。这意味着较大的轮廓将放置在列表的前面，而较小的轮廓将显示在列表的后面。

我们假设我们的考试将成为*图像*的  *主要焦点*，因此比图像中的其他对象更大。这个假设允许我们“过滤”我们的轮廓，只需调查它们的区域并知道对应于检查的轮廓应该在列表的前面附近。

但是，  *轮廓区域和大小*还不够 - 我们还应该检查轮廓上的*顶点*数量  。

为此，我们循环遍历**第40行的**每个（已排序）轮廓  。对于它们中的每一个，我们近似轮廓，这实质上意味着我们  *简化*了轮廓中的点数，使其成为“更基本”的几何形状。您可以在本文中了解有关轮廓近似的更多信息，以[构建移动文档扫描仪。][Link 3]

在  **第47行，**我们检查我们的近似轮廓是否有四个点，如果有，  *我们假设我们已经找到了考试。*

下面我添加了一个示例图像，演示了  在原始图像上绘制的 docCnt变量：

![图3：在原始图像上绘制与检查相关的轮廓的示例，表明我们已成功找到该检查。][3]

**图3：**在原始图像上绘制与检查相关的轮廓的示例，表明我们已成功找到该检查。

果然，这个区域对应于考试大纲。

现在我们已经使用轮廓来查找检查的轮廓，我们可以应用透视变换来获得文档的自上而下的鸟瞰视图：

<div id="crayon-5d52be9f4eb5f463330481" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

<table> 
 <tbody> 
  <tr> 
   <td><br> 
    <div> 
     <div>
       51 
     </div> 
     <div>
       52 
     </div> 
     <div>
       53 
     </div> 
     <div>
       54 
     </div> 
     <div>
       55 
     </div> 
    </div><br> </td> 
   <td> 
    <div> 
     <div> 
      <span># apply a four point perspective transform to both the</span> 
     </div> 
     <div> 
      <span># original image and grayscale image to obtain a top-down</span> 
     </div> 
     <div> 
      <span># birds eye view of the paper</span> 
     </div> 
     <div> 
      <span>paper</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>four_point_transform</span> 
      <span>(</span> 
      <span>image</span> 
      <span>,</span> 
      <span> </span> 
      <span>docCnt</span> 
      <span>.</span> 
      <span>reshape</span> 
      <span>(</span> 
      <span>4</span> 
      <span>,</span> 
      <span> </span> 
      <span>2</span> 
      <span>)</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>warped</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>four_point_transform</span> 
      <span>(</span> 
      <span>gray</span> 
      <span>,</span> 
      <span> </span> 
      <span>docCnt</span> 
      <span>.</span> 
      <span>reshape</span> 
      <span>(</span> 
      <span>4</span> 
      <span>,</span> 
      <span> </span> 
      <span>2</span> 
      <span>)</span> 
      <span>)</span> 
     </div> 
    </div></td> 
  </tr> 
 </tbody> 
</table>

在这种情况下，我们将使用我的 four\_point\_transform   函数的实现：

1.  以*特定的，可重复的方式*对我们轮廓的  *（x，y）*坐标进行排序  *。*
2.  对区域应用透视变换。

您可以在本文中了解有关透视变换的更多[信息][Link 5]  ，以及[有关坐标排序的更新][Link 6]透视变换，但目前只需了解此功能处理“偏斜”考试并对其进行变换，返回自上而下的视图文件：

![图4：获取原始图像和灰度版本的自上而下的鸟瞰图。][4]

**图4：**获得原始图像*（左）*  和灰度版本*（右）*的自上而下鸟瞰图。

好吧，现在我们到了某个地方。

我们在原始图片中找到了考试。

我们应用透视变换来获得文档的90度视角。

但是我们如何对文件进行实际  *评分*呢？

此步骤从*二值化*开始  ，或从  图像*背景*中对*前景*进行阈值处理/分割的过程  ：

<div id="crayon-5d52be9f4eb6a582470897" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

# apply Otsu’s thresholding method to binarize the warped #

# piece of paper #

thresh = cv2.threshold(warped, 0, 255,  
cv2.THRESH\_BINARY\_INV | cv2.THRESH\_OTSU)\[1\]

<table> 
 <tbody> 
  <tr> 
   <td><br> 
    <div> 
     <div>
       57 
     </div> 
     <div>
       58 
     </div> 
     <div>
       59 
     </div> 
     <div>
       60 
     </div> 
    </div><br> </td> 
   <td> 
    <div> 
     <div> 
      <span># apply Otsu’s thresholding method to binarize the warped</span> 
     </div> 
     <div> 
      <span># piece of paper</span> 
     </div> 
     <div> 
      <span>thresh</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>threshold</span> 
      <span>(</span> 
      <span>warped</span> 
      <span>,</span> 
      <span> </span> 
      <span>0</span> 
      <span>,</span> 
      <span> </span> 
      <span>255</span> 
      <span>,</span> 
     </div> 
     <div> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>THRESH_BINARY_INV</span> 
      <span> </span> 
      <span>|</span> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>THRESH_OTSU</span> 
      <span>)</span> 
      <span>[</span> 
      <span>1</span> 
      <span>]</span> 
     </div> 
    </div></td> 
  </tr> 
 </tbody> 
</table>

应用Otsu的阈值处理方法后，我们的考试现在是  *二进制*图像：

![图5：使用Otsu的阈值处理允许我们从图像的背景中分割前景。][5_Otsu]

**图5：**使用Otsu的阈值处理允许我们从图像的背景中分割前景。

注意图像的  *背景 *是  *黑色的*，而  *前景*是  *白色的。*

这种二值化将允许我们再次应用轮廓提取技术来查找考试中的每个气泡：

<div id="crayon-5d52be9f4eb71835583686" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

# apply Otsu’s thresholding method to binarize the warped #

# piece of paper #

thresh = cv2.threshold(warped, 0, 255,  
cv2.THRESH\_BINARY\_INV | cv2.THRESH\_OTSU)\[1\]

# find contours in the thresholded image, then initialize #

# the list of contours that correspond to questions #

cnts = cv2.findContours(thresh.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
questionCnts = \[\]

# loop over the contours #

for c in cnts:  
\# compute the bounding box of the contour, then use the  
\# bounding box to derive the aspect ratio  
(x, y, w, h) = cv2.boundingRect©  
ar = w / float(h)

# in order to label the contour as a question, region
    # should be sufficiently wide, sufficiently tall, and
    # have an aspect ratio approximately equal to 1
    if w >= 20 and h >= 20 and ar >= 0.9 and ar <= 1.1:
    	questionCnts.append(c)</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-62">62</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-63">63</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-64">64</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-65">65</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-66">66</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-67">67</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-68">68</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-69">69</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-70">70</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-71">71</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-72">72</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-73">73</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-74">74</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-75">75</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-76">76</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-77">77</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-78">78</div><div class="crayon-num" data-line="crayon-5d52be9f4eb71835583686-79">79</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4eb71835583686-80">80</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-62"><span class="crayon-c"># find contours in the thresholded image, then initialize</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-63"><span class="crayon-c"># the list of contours that correspond to questions</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-64"><span class="crayon-v">cnts</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">findContours</span><span class="crayon-sy">(</span><span class="crayon-v">thresh</span><span class="crayon-sy">.</span><span class="crayon-k ">copy</span><span class="crayon-sy">(</span><span class="crayon-sy">)</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-v">RETR_EXTERNAL</span><span class="crayon-sy">,</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-65"><span class="crayon-h">	</span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-v">CHAIN_APPROX_SIMPLE</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-66"><span class="crayon-v">cnts</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">imutils</span><span class="crayon-sy">.</span><span class="crayon-e">grab_contours</span><span class="crayon-sy">(</span><span class="crayon-v">cnts</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-67"><span class="crayon-v">questionCnts</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-sy">[</span><span class="crayon-sy">]</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-68"> </div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-69"><span class="crayon-c"># loop over the contours</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-70"><span class="crayon-st">for</span><span class="crayon-h"> </span><span class="crayon-i">c</span><span class="crayon-h"> </span><span class="crayon-st">in</span><span class="crayon-h"> </span><span class="crayon-v">cnts</span><span class="crayon-o">:</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-71"><span class="crayon-h">	</span><span class="crayon-c"># compute the bounding box of the contour, then use the</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-72"><span class="crayon-h">	</span><span class="crayon-c"># bounding box to derive the aspect ratio</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-73"><span class="crayon-h">	</span><span class="crayon-sy">(</span><span class="crayon-v">x</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">y</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-k ">w</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">h</span><span class="crayon-sy">)</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">boundingRect</span><span class="crayon-sy">(</span><span class="crayon-v">c</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-74"><span class="crayon-h">	</span><span class="crayon-v">ar</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-k ">w</span><span class="crayon-h"> </span><span class="crayon-o">/</span><span class="crayon-h"> </span><span class="crayon-k ">float</span><span class="crayon-sy">(</span><span class="crayon-v">h</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-75"> </div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-76"><span class="crayon-h">	</span><span class="crayon-c"># in order to label the contour as a question, region</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-77"><span class="crayon-h">	</span><span class="crayon-c"># should be sufficiently wide, sufficiently tall, and</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-78"><span class="crayon-h">	</span><span class="crayon-c"># have an aspect ratio approximately equal to 1</span></div><div class="crayon-line" id="crayon-5d52be9f4eb71835583686-79"><span class="crayon-h">	</span><span class="crayon-st">if</span><span class="crayon-h"> </span><span class="crayon-k ">w</span><span class="crayon-h"> </span><span class="crayon-o">>=</span><span class="crayon-h"> </span><span class="crayon-cn">20</span><span class="crayon-h"> </span><span class="crayon-st">and</span><span class="crayon-h"> </span><span class="crayon-v">h</span><span class="crayon-h"> </span><span class="crayon-o">>=</span><span class="crayon-h"> </span><span class="crayon-cn">20</span><span class="crayon-h"> </span><span class="crayon-st">and</span><span class="crayon-h"> </span><span class="crayon-v">ar</span><span class="crayon-h"> </span><span class="crayon-o">>=</span><span class="crayon-h"> </span><span class="crayon-cn">0.9</span><span class="crayon-h"> </span><span class="crayon-st">and</span><span class="crayon-h"> </span><span class="crayon-v">ar</span><span class="crayon-h"> </span><span class="crayon-o"><=</span><span class="crayon-h"> </span><span class="crayon-cn">1.1</span><span class="crayon-o">:</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4eb71835583686-80"><span class="crayon-h">		</span><span class="crayon-v">questionCnts</span><span class="crayon-sy">.</span><span class="crayon-e">append</span><span class="crayon-sy">(</span><span class="crayon-v">c</span><span class="crayon-sy">)</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

**第64-67行**处理在我们的thresh   二进制图像上找到轮廓 ，然后初始化 questionCnts  ，这是一个与检查中的问题/气泡相对应的轮廓列表。

为了确定图像的哪些区域是气泡，我们首先遍历每个单独的轮廓（**第70行**）。

对于这些轮廓中的每一个，我们计算边界框（**第73行**），这也允许我们计算*纵横比*，或者更简单地，计算宽度与高度的比率（**第74行**）。

为了将轮廓区域视为气泡，该区域应：

1.  足够宽和高（在这种情况下，两个维度中至少20个像素）。
2.  宽高比*大约*等于1。

只要这些检查成立，我们就可以更新我们的 questionCnts   列表并将该区域标记为气泡。

下面我已经包含了一个截图，它  在我们的图像上绘制了questionCnts的输出 ：

![图6：使用轮廓过滤可以让我们在气泡纸检查识别软件中找到所有问题气泡。][6]

**图6：**使用轮廓过滤可以让我们在气泡纸检查识别软件中找到所有问题气泡。

请注意如何  *仅*突出显示考试的问题区域，而不是其他任何内容。

我们现在可以转到我们的OMR系统的“评级”部分：

<div id="crayon-5d52be9f4eb89918688243" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

# apply Otsu’s thresholding method to binarize the warped #

# piece of paper #

thresh = cv2.threshold(warped, 0, 255,  
cv2.THRESH\_BINARY\_INV | cv2.THRESH\_OTSU)\[1\]

# find contours in the thresholded image, then initialize #

# the list of contours that correspond to questions #

cnts = cv2.findContours(thresh.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
questionCnts = \[\]

# loop over the contours #

for c in cnts:  
\# compute the bounding box of the contour, then use the  
\# bounding box to derive the aspect ratio  
(x, y, w, h) = cv2.boundingRect©  
ar = w / float(h)

# sort the question contours top-to-bottom, then initialize #

# the total number of correct answers #

questionCnts = contours.sort\_contours(questionCnts,  
method=“top-to-bottom”)\[0\]  
correct = 0

# each question has 5 possible answers, to loop over the #

# question in batches of 5 #

for (q, i) in enumerate(np.arange(0, len(questionCnts), 5)):  
\# sort the contours for the current question from  
\# left to right, then initialize the index of the  
\# bubbled answer  
cnts = contours.sort\_contours(questionCnts\[i:i + 5\])\[0\]  
bubbled = None

<table> 
 <tbody> 
  <tr> 
   <td><br> 
    <div> 
     <div>
       82 
     </div> 
     <div>
       83 
     </div> 
     <div>
       84 
     </div> 
     <div>
       85 
     </div> 
     <div>
       86 
     </div> 
     <div>
       87 
     </div> 
     <div>
       88 
     </div> 
     <div>
       89 
     </div> 
     <div>
       90 
     </div> 
     <div>
       91 
     </div> 
     <div>
       92 
     </div> 
     <div>
       93 
     </div> 
     <div>
       94 
     </div> 
     <div>
       95 
     </div> 
    </div><br> </td> 
   <td> 
    <div> 
     <div> 
      <span># sort the question contours top-to-bottom, then initialize</span> 
     </div> 
     <div> 
      <span># the total number of correct answers</span> 
     </div> 
     <div> 
      <span>questionCnts</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>contours</span> 
      <span>.</span> 
      <span>sort_contours</span> 
      <span>(</span> 
      <span>questionCnts</span> 
      <span>,</span> 
     </div> 
     <div> 
      <span> </span> 
      <span>method</span> 
      <span>=</span> 
      <span>“top-to-bottom”</span> 
      <span>)</span> 
      <span>[</span> 
      <span>0</span> 
      <span>]</span> 
     </div> 
     <div> 
      <span>correct</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>0</span> 
     </div> 
     <div>
       &nbsp; 
     </div> 
     <div> 
      <span># each question has 5 possible answers, to loop over the</span> 
     </div> 
     <div> 
      <span># question in batches of 5</span> 
     </div> 
     <div> 
      <span>for</span> 
      <span> </span> 
      <span>(</span> 
      <span>q</span> 
      <span>,</span> 
      <span> </span> 
      <span>i</span> 
      <span>)</span> 
      <span> </span> 
      <span>in</span> 
      <span> </span> 
      <span>enumerate</span> 
      <span>(</span> 
      <span>np</span> 
      <span>.</span> 
      <span>arange</span> 
      <span>(</span> 
      <span>0</span> 
      <span>,</span> 
      <span> </span> 
      <span>len</span> 
      <span>(</span> 
      <span>questionCnts</span> 
      <span>)</span> 
      <span>,</span> 
      <span> </span> 
      <span>5</span> 
      <span>)</span> 
      <span>)</span> 
      <span>:</span> 
     </div> 
     <div> 
      <span> </span> 
      <span># sort the contours for the current question from</span> 
     </div> 
     <div> 
      <span> </span> 
      <span># left to right, then initialize the index of the</span> 
     </div> 
     <div> 
      <span> </span> 
      <span># bubbled answer</span> 
     </div> 
     <div> 
      <span> </span> 
      <span>cnts</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>contours</span> 
      <span>.</span> 
      <span>sort_contours</span> 
      <span>(</span> 
      <span>questionCnts</span> 
      <span>[</span> 
      <span>i</span> 
      <span>:</span> 
      <span>i</span> 
      <span> </span> 
      <span>+</span> 
      <span> </span> 
      <span>5</span> 
      <span>]</span> 
      <span>)</span> 
      <span>[</span> 
      <span>0</span> 
      <span>]</span> 
     </div> 
     <div> 
      <span> </span> 
      <span>bubbled</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>None</span> 
     </div> 
    </div></td> 
  </tr> 
 </tbody> 
</table>

首先，我们必须从头到尾排序我们的 问题   。这将确保*更接近*考试*顶部*的问题行将  *首先*出现在排序列表中。

我们还初始化一个簿记变量来跟踪正确   答案的数量 。

在  **第90行，**我们开始循环我们的问题。由于每个问题都有5个可能的答案，我们将应用NumPy数组切片和轮廓排序来从左到右对*当前轮廓组*进行排序  。

这种方法的工作原因是因为我们*已经*从上到下对轮廓进行  *了*排序。我们  *知道*每个问题的5个气泡将按顺序出现在我们的列表中 -  **但我们  *不知道*  这些气泡是否会从左到右排序。 **第**94行**上的排序轮廓调用  负责处理此问题，并确保每行轮廓按从左到右的顺序排列。

为了形象化这个概念，我在下面列出了一个截图，它将每行问题描述为一个单独的颜色：

![图7：通过从上到下排序我们的轮廓，然后从左到右排序，我们可以提取每一行气泡。 因此，每行等于一个问题的气泡。][7_]

**图7：**通过从上到下排序我们的轮廓，然后从左到右排序，我们可以提取每一行气泡。因此，每行等于一个问题的气泡。

给定一排气泡，下一步是确定填充了哪个气泡。

我们可以通过使用我们的阈值   图像并计算每个气泡区域中的非零像素（即*前景像素*）的数量  来实现这一点 ：

<div id="crayon-5d52be9f4ebc2488133240" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

# apply Otsu’s thresholding method to binarize the warped #

# piece of paper #

thresh = cv2.threshold(warped, 0, 255,  
cv2.THRESH\_BINARY\_INV | cv2.THRESH\_OTSU)\[1\]

# find contours in the thresholded image, then initialize #

# the list of contours that correspond to questions #

cnts = cv2.findContours(thresh.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
questionCnts = \[\]

# loop over the contours #

for c in cnts:  
\# compute the bounding box of the contour, then use the  
\# bounding box to derive the aspect ratio  
(x, y, w, h) = cv2.boundingRect©  
ar = w / float(h)

# sort the question contours top-to-bottom, then initialize #

# the total number of correct answers #

questionCnts = contours.sort\_contours(questionCnts,  
method=“top-to-bottom”)\[0\]  
correct = 0

# each question has 5 possible answers, to loop over the #

# question in batches of 5 #

# loop over the sorted contours
    for (j, c) in enumerate(cnts):
    	# construct a mask that reveals only the current
    	# "bubble" for the question
    	mask = np.zeros(thresh.shape, dtype="uint8")
    	cv2.drawContours(mask, [c], -1, 255, -1)
    
    	# apply the mask to the thresholded image, then
    	# count the number of non-zero pixels in the
    	# bubble area
    	mask = cv2.bitwise_and(thresh, thresh, mask=mask)
    	total = cv2.countNonZero(mask)
    
    	# if the current total has a larger number of total
    	# non-zero pixels, then we are examining the currently
    	# bubbled-in answer
    	if bubbled is None or total > bubbled[0]:
    		bubbled = (total, j)</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-97">97</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-98">98</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-99">99</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-100">100</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-101">101</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-102">102</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-103">103</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-104">104</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-105">105</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-106">106</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-107">107</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-108">108</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-109">109</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-110">110</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-111">111</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-112">112</div><div class="crayon-num" data-line="crayon-5d52be9f4ebc2488133240-113">113</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebc2488133240-114">114</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-97"><span class="crayon-h">	</span><span class="crayon-c"># loop over the sorted contours</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-98"><span class="crayon-h">	</span><span class="crayon-st">for</span><span class="crayon-h"> </span><span class="crayon-sy">(</span><span class="crayon-v">j</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">c</span><span class="crayon-sy">)</span><span class="crayon-h"> </span><span class="crayon-st">in</span><span class="crayon-h"> </span><span class="crayon-k ">enumerate</span><span class="crayon-sy">(</span><span class="crayon-v">cnts</span><span class="crayon-sy">)</span><span class="crayon-o">:</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-99"><span class="crayon-h">		</span><span class="crayon-c"># construct a mask that reveals only the current</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-100"><span class="crayon-h">		</span><span class="crayon-c"># "bubble" for the question</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-101"><span class="crayon-h">		</span><span class="crayon-v">mask</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">np</span><span class="crayon-sy">.</span><span class="crayon-e">zeros</span><span class="crayon-sy">(</span><span class="crayon-v">thresh</span><span class="crayon-sy">.</span><span class="crayon-v">shape</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">dtype</span><span class="crayon-o">=</span><span class="crayon-s">"uint8"</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-102"><span class="crayon-h">		</span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">drawContours</span><span class="crayon-sy">(</span><span class="crayon-v">mask</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-sy">[</span><span class="crayon-v">c</span><span class="crayon-sy">]</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-o">-</span><span class="crayon-cn">1</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">255</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-o">-</span><span class="crayon-cn">1</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-103"> </div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-104"><span class="crayon-h">		</span><span class="crayon-c"># apply the mask to the thresholded image, then</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-105"><span class="crayon-h">		</span><span class="crayon-c"># count the number of non-zero pixels in the</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-106"><span class="crayon-h">		</span><span class="crayon-c"># bubble area</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-107"><span class="crayon-h">		</span><span class="crayon-v">mask</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">bitwise_and</span><span class="crayon-sy">(</span><span class="crayon-v">thresh</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">thresh</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">mask</span><span class="crayon-o">=</span><span class="crayon-v">mask</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-108"><span class="crayon-h">		</span><span class="crayon-v">total</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">countNonZero</span><span class="crayon-sy">(</span><span class="crayon-v">mask</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-109"> </div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-110"><span class="crayon-h">		</span><span class="crayon-c"># if the current total has a larger number of total</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-111"><span class="crayon-h">		</span><span class="crayon-c"># non-zero pixels, then we are examining the currently</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-112"><span class="crayon-h">		</span><span class="crayon-c"># bubbled-in answer</span></div><div class="crayon-line" id="crayon-5d52be9f4ebc2488133240-113"><span class="crayon-h">		</span><span class="crayon-st">if</span><span class="crayon-h"> </span><span class="crayon-e">bubbled </span><span class="crayon-st">is</span><span class="crayon-h"> </span><span class="crayon-t">None</span><span class="crayon-h"> </span><span class="crayon-st">or</span><span class="crayon-h"> </span><span class="crayon-v">total</span><span class="crayon-h"> </span><span class="crayon-o">></span><span class="crayon-h"> </span><span class="crayon-v">bubbled</span><span class="crayon-sy">[</span><span class="crayon-cn">0</span><span class="crayon-sy">]</span><span class="crayon-o">:</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebc2488133240-114"><span class="crayon-h">			</span><span class="crayon-v">bubbled</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-sy">(</span><span class="crayon-v">total</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">j</span><span class="crayon-sy">)</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

**第98行**处理**行中**每个已排序气泡的循环。

然后，我们在**101行**上为当前气泡构建一个掩码  ，然后计算掩码区域中非零像素的数量（**第107和108行**）。我们计算的非零像素越多，那么前景像素越多，因此具有最大非零计数的气泡是测试者冒泡的气泡的索引（**第113和114行**）。

下面我列举了一个为与问题相关的每个气泡创建和应用蒙版的示例：

![图8：为一行中的每个气泡构建掩码的示例。][8]

**图8：**为一行中的每个气泡构建掩码的示例。

显然，与*“B”*相关联的气泡  具有最多阈值像素，因此是用户在他们的考试中标记的气泡。

下一个代码块处理在ANSWER\_KEY中查找正确答案  ，更新任何相关的簿记员变量，最后在我们的图像上绘制标记的气泡：

<div id="crayon-5d52be9f4ebd0582812466" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

# apply Otsu’s thresholding method to binarize the warped #

# piece of paper #

thresh = cv2.threshold(warped, 0, 255,  
cv2.THRESH\_BINARY\_INV | cv2.THRESH\_OTSU)\[1\]

# find contours in the thresholded image, then initialize #

# the list of contours that correspond to questions #

cnts = cv2.findContours(thresh.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
questionCnts = \[\]

# loop over the contours #

for c in cnts:  
\# compute the bounding box of the contour, then use the  
\# bounding box to derive the aspect ratio  
(x, y, w, h) = cv2.boundingRect©  
ar = w / float(h)

# sort the question contours top-to-bottom, then initialize #

# the total number of correct answers #

questionCnts = contours.sort\_contours(questionCnts,  
method=“top-to-bottom”)\[0\]  
correct = 0

# each question has 5 possible answers, to loop over the #

# question in batches of 5 #

# loop over the sorted contours
    for (j, c) in enumerate(cnts):
    	# construct a mask that reveals only the current
    	# "bubble" for the question
    	mask = np.zeros(thresh.shape, dtype="uint8")
    	cv2.drawContours(mask, [c], -1, 255, -1)
    
    	# apply the mask to the thresholded image, then
    	# count the number of non-zero pixels in the
    	# bubble area
    	mask = cv2.bitwise_and(thresh, thresh, mask=mask)
    	total = cv2.countNonZero(mask)
    
    	# if the current total has a larger number of total
    	# non-zero pixels, then we are examining the currently
    	# bubbled-in answer
    	if bubbled is None or total > bubbled[0]:
    		bubbled = (total, j)
    
    # initialize the contour color and the index of the
    # *correct* answer
    color = (0, 0, 255)
    k = ANSWER_KEY[q]
    
    # check to see if the bubbled answer is correct
    if k == bubbled[1]:
    	color = (0, 255, 0)
    	correct += 1
    
    # draw the outline of the correct answer on the test
    cv2.drawContours(paper, [cnts[k]], -1, color, 3)</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebd0582812466-116">116</div><div class="crayon-num" data-line="crayon-5d52be9f4ebd0582812466-117">117</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebd0582812466-118">118</div><div class="crayon-num" data-line="crayon-5d52be9f4ebd0582812466-119">119</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebd0582812466-120">120</div><div class="crayon-num" data-line="crayon-5d52be9f4ebd0582812466-121">121</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebd0582812466-122">122</div><div class="crayon-num" data-line="crayon-5d52be9f4ebd0582812466-123">123</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebd0582812466-124">124</div><div class="crayon-num" data-line="crayon-5d52be9f4ebd0582812466-125">125</div><div class="crayon-num crayon-striped-num" data-line="crayon-5d52be9f4ebd0582812466-126">126</div><div class="crayon-num" data-line="crayon-5d52be9f4ebd0582812466-127">127</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebd0582812466-116"><span class="crayon-h">	</span><span class="crayon-c"># initialize the contour color and the index of the</span></div><div class="crayon-line" id="crayon-5d52be9f4ebd0582812466-117"><span class="crayon-h">	</span><span class="crayon-c"># *correct* answer</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebd0582812466-118"><span class="crayon-h">	</span><span class="crayon-v">color</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-sy">(</span><span class="crayon-cn">0</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">0</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">255</span><span class="crayon-sy">)</span></div><div class="crayon-line" id="crayon-5d52be9f4ebd0582812466-119"><span class="crayon-h">	</span><span class="crayon-v">k</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-v">ANSWER_KEY</span><span class="crayon-sy">[</span><span class="crayon-v">q</span><span class="crayon-sy">]</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebd0582812466-120"> </div><div class="crayon-line" id="crayon-5d52be9f4ebd0582812466-121"><span class="crayon-h">	</span><span class="crayon-c"># check to see if the bubbled answer is correct</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebd0582812466-122"><span class="crayon-h">	</span><span class="crayon-st">if</span><span class="crayon-h"> </span><span class="crayon-v">k</span><span class="crayon-h"> </span><span class="crayon-o">==</span><span class="crayon-h"> </span><span class="crayon-v">bubbled</span><span class="crayon-sy">[</span><span class="crayon-cn">1</span><span class="crayon-sy">]</span><span class="crayon-o">:</span></div><div class="crayon-line" id="crayon-5d52be9f4ebd0582812466-123"><span class="crayon-h">		</span><span class="crayon-v">color</span><span class="crayon-h"> </span><span class="crayon-o">=</span><span class="crayon-h"> </span><span class="crayon-sy">(</span><span class="crayon-cn">0</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">255</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">0</span><span class="crayon-sy">)</span></div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebd0582812466-124"><span class="crayon-h">		</span><span class="crayon-v">correct</span><span class="crayon-h"> </span><span class="crayon-o">+=</span><span class="crayon-h"> </span><span class="crayon-cn">1</span></div><div class="crayon-line" id="crayon-5d52be9f4ebd0582812466-125"> </div><div class="crayon-line crayon-striped-line" id="crayon-5d52be9f4ebd0582812466-126"><span class="crayon-h">	</span><span class="crayon-c"># draw the outline of the correct answer on the test</span></div><div class="crayon-line" id="crayon-5d52be9f4ebd0582812466-127"><span class="crayon-h">	</span><span class="crayon-v">cv2</span><span class="crayon-sy">.</span><span class="crayon-e">drawContours</span><span class="crayon-sy">(</span><span class="crayon-v">paper</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-sy">[</span><span class="crayon-v">cnts</span><span class="crayon-sy">[</span><span class="crayon-v">k</span><span class="crayon-sy">]</span><span class="crayon-sy">]</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-o">-</span><span class="crayon-cn">1</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-v">color</span><span class="crayon-sy">,</span><span class="crayon-h"> </span><span class="crayon-cn">3</span><span class="crayon-sy">)</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

根据考生是否正确或不正确，得出考试中的颜色。如果考生是  *正确的*，我们将以*绿色*突出显示他们的答案  。但是，如果考生犯了错误并且标记了错误答案，我们会通过用*红色*突出显示*正确*答案来  告诉他们  ：

![图9：绘制“绿色”圆圈以标记“正确”或“红色”圆圈以标记“不正确”。][9]

**图9：**绘制“绿色”圆圈以标记“正确”或“红色”圆圈以标记“不正确”。

最后，我们的最后一个代码块处理考试评分并将结果显示在我们的屏幕上：

<div id="crayon-5d52be9f4ebd9599305437" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Python</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;"># import the necessary packages

from imutils.perspective import four\_point\_transform  
from imutils import contours  
import numpy as np  
import argparse  
import imutils  
import cv2

# construct the argument parse and parse the arguments #

ap = argparse.ArgumentParser()  
ap.add\_argument("-i", “–image”, required=True,  
help=“path to the input image”)  
args = vars(ap.parse\_args())

# define the answer key which maps the question number #

# to the correct answer #

ANSWER\_KEY = \{0: 1, 1: 4, 2: 0, 3: 3, 4: 1\}

# load the image, convert it to grayscale, blur it #

# slightly, then find edges #

image = cv2.imread(args\[“image”\])  
gray = cv2.cvtColor(image, cv2.COLOR\_BGR2GRAY)  
blurred = cv2.GaussianBlur(gray, (5, 5), 0)  
edged = cv2.Canny(blurred, 75, 200)

# find contours in the edge map, then initialize #

# the contour that corresponds to the document #

cnts = cv2.findContours(edged.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
docCnt = None

# ensure that at least one contour was found #

if len(cnts) > 0:  
\# sort the contours according to their size in  
\# descending order  
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)

# apply a four point perspective transform to both the #

# original image and grayscale image to obtain a top-down #

# birds eye view of the paper #

paper = four\_point\_transform(image, docCnt.reshape(4, 2))  
warped = four\_point\_transform(gray, docCnt.reshape(4, 2))

# apply Otsu’s thresholding method to binarize the warped #

# piece of paper #

thresh = cv2.threshold(warped, 0, 255,  
cv2.THRESH\_BINARY\_INV | cv2.THRESH\_OTSU)\[1\]

# find contours in the thresholded image, then initialize #

# the list of contours that correspond to questions #

cnts = cv2.findContours(thresh.copy(), cv2.RETR\_EXTERNAL,  
cv2.CHAIN\_APPROX\_SIMPLE)  
cnts = imutils.grab\_contours(cnts)  
questionCnts = \[\]

# loop over the contours #

for c in cnts:  
\# compute the bounding box of the contour, then use the  
\# bounding box to derive the aspect ratio  
(x, y, w, h) = cv2.boundingRect©  
ar = w / float(h)

# sort the question contours top-to-bottom, then initialize #

# the total number of correct answers #

questionCnts = contours.sort\_contours(questionCnts,  
method=“top-to-bottom”)\[0\]  
correct = 0

# each question has 5 possible answers, to loop over the #

# question in batches of 5 #

# grab the test taker #

score = (correct / 5.0) \* 100  
print("\[INFO\] score: \{:.2f\}%".format(score))  
cv2.putText(paper, “\{:.2f\}%”.format(score), (10, 30),  
cv2.FONT\_HERSHEY\_SIMPLEX, 0.9, (0, 0, 255), 2)  
cv2.imshow(“Original”, image)  
cv2.imshow(“Exam”, paper)  
cv2.waitKey(0)

<table> 
 <tbody> 
  <tr> 
   <td><br> 
    <div> 
     <div>
       129 
     </div> 
     <div>
       130 
     </div> 
     <div>
       131 
     </div> 
     <div>
       132 
     </div> 
     <div>
       133 
     </div> 
     <div>
       134 
     </div> 
     <div>
       135 
     </div> 
     <div>
       136 
     </div> 
    </div><br> </td> 
   <td> 
    <div> 
     <div> 
      <span># grab the test taker</span> 
     </div> 
     <div> 
      <span>score</span> 
      <span> </span> 
      <span>=</span> 
      <span> </span> 
      <span>(</span> 
      <span>correct</span> 
      <span> </span> 
      <span>/</span> 
      <span> </span> 
      <span>5.0</span> 
      <span>)</span> 
      <span> </span> 
      <span>*</span> 
      <span> </span> 
      <span>100</span> 
     </div> 
     <div> 
      <span>print</span> 
      <span>(</span> 
      <span>"[INFO] score: {:.2f}%"</span> 
      <span>.</span> 
      <span>format</span> 
      <span>(</span> 
      <span>score</span> 
      <span>)</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>cv2</span> 
      <span>.</span> 
      <span>putText</span> 
      <span>(</span> 
      <span>paper</span> 
      <span>,</span> 
      <span> </span> 
      <span>“{:.2f}%”</span> 
      <span>.</span> 
      <span>format</span> 
      <span>(</span> 
      <span>score</span> 
      <span>)</span> 
      <span>,</span> 
      <span> </span> 
      <span>(</span> 
      <span>10</span> 
      <span>,</span> 
      <span> </span> 
      <span>30</span> 
      <span>)</span> 
      <span>,</span> 
     </div> 
     <div> 
      <span> </span> 
      <span>cv2</span> 
      <span>.</span> 
      <span>FONT_HERSHEY_SIMPLEX</span> 
      <span>,</span> 
      <span> </span> 
      <span>0.9</span> 
      <span>,</span> 
      <span> </span> 
      <span>(</span> 
      <span>0</span> 
      <span>,</span> 
      <span> </span> 
      <span>0</span> 
      <span>,</span> 
      <span> </span> 
      <span>255</span> 
      <span>)</span> 
      <span>,</span> 
      <span> </span> 
      <span>2</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>cv2</span> 
      <span>.</span> 
      <span>imshow</span> 
      <span>(</span> 
      <span>“Original”</span> 
      <span>,</span> 
      <span> </span> 
      <span>image</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>cv2</span> 
      <span>.</span> 
      <span>imshow</span> 
      <span>(</span> 
      <span>“Exam”</span> 
      <span>,</span> 
      <span> </span> 
      <span>paper</span> 
      <span>)</span> 
     </div> 
     <div> 
      <span>cv2</span> 
      <span>.</span> 
      <span>waitKey</span> 
      <span>(</span> 
      <span>0</span> 
      <span>)</span> 
     </div> 
    </div></td> 
  </tr> 
 </tbody> 
</table>

您可以在下面看到我们完全评分的示例图像的输出：

![图10：完成我们的OMR系统，用于对人工考试进行评分。][10_OMR]

**图10：**完成我们的OMR系统，用于对人工考试进行评分。

在这种情况下，读者获得了80％的考试成绩。他们错过的唯一问题是＃4，他们错误地将*“C”*标记  为正确答案（*“D”*是正确的选择）。

### 为什么不使用圆检测？ ###

完成本教程后，您可能想知道：

*“嘿阿德里安，答案泡沫是一个圆圈。那你为什么要提取轮廓而不是应用[霍夫圆][Link 7]来找到图像中的圆圈呢？“*

好问题。

首先，在图像到图像的基础上将参数调整为霍夫圆可能是一个真正的痛苦。但这只是一个小原因。

的  **真正原因**是：

***用户错误。***

有多少次，无论是否故意，你是否已经填写了泡泡纸上的线条？我不是专家，但我必须猜测，每20分中至少有1分，考生填写的是“略微”在线外。

你猜怎么着？

霍夫圆圈不能很好地处理轮廓中的变形 - 在这种情况下，圆检测将完全失败。

因此，我建议使用轮廓和轮廓属性来帮助您过滤气泡和答案。该 CV2 。findContours   功能并不关心泡沫是“圆形”，“完美圆形”，还是“天哪，这到底是什么？”。

相反， cv2 。findContours   函数将返回一组  *blob*给你，它将是你图像中的前景区域。然后，您可以对这些区域进行处理并对其进行过滤以找到您的问题（正如我们在本教程中所做的那样），并按照您的方式进行操作。

### 我们的气泡板测试扫描仪和分级机结果 ###

要查看我们的气泡表测试分级器，请务必使用本教程底部的***“下载”***部分将源代码和示例图像***下载***到此帖子  。

我们已经看过 test\_01了。png   作为本文前面的例子，让我们试试 test\_02 。png  ：

<div id="crayon-5d52be9f4ebf5395739483" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Shell</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;">$ python test_grader.py --image images/test_02.png</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num" data-line="crayon-5d52be9f4ebf5395739483-1">1</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line" id="crayon-5d52be9f4ebf5395739483-1"><span class="crayon-sy">$</span><span class="crayon-h"> </span><span class="crayon-e">python </span><span class="crayon-v">test_grader</span><span class="crayon-e">.py</span><span class="crayon-h"> </span><span class="crayon-o">--</span><span class="crayon-e">image </span><span class="crayon-v">images</span><span class="crayon-o">/</span><span class="crayon-v">test_02</span><span class="crayon-e">.png</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

在这里，我们可以看到一个特别邪恶的用户参加了我们的考试。他们对测试不满意，在其前面写着*“\#yourtestsux”*以及一个无政府状态鼓舞人心的  *“\#breakthesystem”*。他们还为所有答案标记为  *“A”*。

也许毫不奇怪，用户在考试中获得了可怜的20％，完全基于运气：

![图11：通过使用轮廓过滤，我们可以忽略检查区域，否则会损害其完整性。][11]

**图11：**通过使用轮廓过滤，我们可以忽略检查区域，否则会损害其完整性。

我们试试另一张图片：

<div id="crayon-5d52be9f4ebfa907262418" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Shell</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;">$ python test_grader.py --image images/test_03.png</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num" data-line="crayon-5d52be9f4ebfa907262418-1">1</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line" id="crayon-5d52be9f4ebfa907262418-1"><span class="crayon-sy">$</span><span class="crayon-h"> </span><span class="crayon-e">python </span><span class="crayon-v">test_grader</span><span class="crayon-e">.py</span><span class="crayon-h"> </span><span class="crayon-o">--</span><span class="crayon-e">image </span><span class="crayon-v">images</span><span class="crayon-o">/</span><span class="crayon-v">test_03</span><span class="crayon-e">.png</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

这次读者做得好一点，得分为60％：

![图12：使用Python和OpenCV构建气泡板扫描仪和测试分级机。][12_Python_OpenCV]

**图12：**使用Python和OpenCV构建气泡板扫描仪和测试分级机。

在这个特定的例子中，读者只是沿着对角线标记了所有答案：

<div id="crayon-5d52be9f4ebff307619276" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Shell</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;">$ python test_grader.py --image images/test_04.png</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num" data-line="crayon-5d52be9f4ebff307619276-1">1</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line" id="crayon-5d52be9f4ebff307619276-1"><span class="crayon-sy">$</span><span class="crayon-h"> </span><span class="crayon-e">python </span><span class="crayon-v">test_grader</span><span class="crayon-e">.py</span><span class="crayon-h"> </span><span class="crayon-o">--</span><span class="crayon-e">image </span><span class="crayon-v">images</span><span class="crayon-o">/</span><span class="crayon-v">test_04</span><span class="crayon-e">.png</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

![图13：使用Python和OpenCV进行测试评分的光学标记识别。][13_Python_OpenCV]

**图13：**使用Python和OpenCV进行测试评分的光学标记识别。

不幸的是，对于考生而言，这种策略并没有得到很好的回报。

让我们看一个最后的例子：

<div id="crayon-5d52be9f4ec04894098818" class="crayon-syntax crayon-theme-classic crayon-font-monaco crayon-os-pc print-yes notranslate" data-settings=" minimize scroll-mouseover" style="margin-top: 12px; margin-bottom: 12px; font-size: 12px !important; line-height: 15px !important; height: auto;">
    	
    		<div class="crayon-toolbar" data-settings=" show" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><span class="crayon-title">Bubble sheet scanner and test grader using OMR, Python and OpenCV</span>
    		<div class="crayon-tools" style="font-size: 12px !important;height: 18px !important; line-height: 18px !important;"><div class="crayon-button crayon-nums-button crayon-pressed" title="Toggle Line Numbers"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-plain-button" title="Toggle Plain Code"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-wrap-button" title="Toggle Line Wrap"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-expand-button" title="Expand Code" style="display: none;"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-copy-button" title="Copy"><div class="crayon-button-icon"></div></div><div class="crayon-button crayon-popup-button" title="Open Code In New Window"><div class="crayon-button-icon"></div></div><span class="crayon-language">Shell</span></div></div>
    		<div class="crayon-info" style="min-height: 16.8px !important; line-height: 16.8px !important;"></div>
    		<div class="crayon-plain-wrap"><textarea wrap="soft" class="crayon-plain print-no" data-settings="" readonly="" style="tab-size: 4; font-size: 12px !important; line-height: 15px !important; z-index: 0; opacity: 0; overflow: hidden;">$ python test_grader.py --image images/test_05.png</textarea></div>
    		<div class="crayon-main" style="position: relative; z-index: 1; overflow: hidden;">
    			<table class="crayon-table" style="">
    				<tbody><tr class="crayon-row">
    			<td class="crayon-nums " data-settings="show">
    				<div class="crayon-nums-content" style="font-size: 12px !important; line-height: 15px !important;"><div class="crayon-num" data-line="crayon-5d52be9f4ec04894098818-1">1</div></div>
    			</td>
    					<td class="crayon-code"><div class="crayon-pre" style="font-size: 12px !important; line-height: 15px !important; -moz-tab-size:4; -o-tab-size:4; -webkit-tab-size:4; tab-size:4;"><div class="crayon-line" id="crayon-5d52be9f4ec04894098818-1"><span class="crayon-sy">$</span><span class="crayon-h"> </span><span class="crayon-e">python </span><span class="crayon-v">test_grader</span><span class="crayon-e">.py</span><span class="crayon-h"> </span><span class="crayon-o">--</span><span class="crayon-e">image </span><span class="crayon-v">images</span><span class="crayon-o">/</span><span class="crayon-v">test_05</span><span class="crayon-e">.png</span></div></div></td>
    				</tr>
    			</tbody></table>
    		</div>
    	</div>

![图14：使用计算机视觉识别气泡板检查。][14]

**图14：**使用计算机视觉识别气泡板检查。

这名学生提前清楚地学习，在考试中获得完美的100％。

### 扩展OMR和测试扫描仪 ###

不可否认，今年夏天/初秋  是我生命中*最忙碌的*时期之一，所以我需要将[OMR][]的开发和测试扫描仪软件的时间安排在上周五的一个缩短的下午。

虽然我能够获得实施的*工作*气泡纸测试扫描仪的准系统  ，但肯定有一些领域需要改进。最明显的改进领域是*处理非填充气泡*的  *逻辑。*

在当前的实现中，我们（天真地）假设读者已经在每个问题行中填充了  *一个*且  *仅一个*气泡。

但是，由于我们仅通过计算一行中阈值像素的数量然后按降序排序来确定特定气泡是否“填充”，这可能会导致两个问题：

1.  如果用户*没有*  在特定问题的答案中冒泡，会发生什么  ？
2.  如果用户是邪恶的并且在同一行中将*多个*气泡标记  为“正确”怎么办？

幸运的是，检测和处理这些问题并不是非常具有挑战性，我们只需要插入一些逻辑。

对于问题＃1，如果读者选择  *不在*特定行的答案中冒泡，那么我们可以在  **第108**行放置  *最小阈值*，我们计算 cv2 。countNonZero  ：

![图15：检测用户是否在考试中标记了零气泡。][15]

**图15：**检测用户是否在考试中标记了零气泡。

如果此值足够大，那么我们可以将气泡标记为“填充”。相反，如果 总数   太小，那么我们可以跳过那个特定的泡沫。如果在行的末尾  *没有 *具有足够大的阈值计数的气泡，我们可以将该问题标记为由测试者“跳过”。

类似的一组步骤可以应用于问题＃2，其中用户将*多个*气泡标记  为单个问题的正确：

![图16：检测用户是否为给定问题标记了多个气泡。][16]

**图16：**检测用户是否为给定问题标记了多个气泡。

同样，我们需要做的就是应用我们的阈值和计数步骤，这次跟踪是否有  *多个气泡*的 总数   超过某个预定值。如果是这样，我们可以使问题无效并将问题标记为不正确。

## 摘要 ##

在这篇博客文章中，我演示了如何使用计算机视觉和图像处理技术构建气泡纸扫描仪和测试分级机。

具体来说，我们实施了  *光学标记识别 *（OMR）方法，这些方法有助于我们捕获人工标记文档并  *自动*分析结果。

最后，我提供了一个Python和OpenCV实现，您可以使用它来构建自己的气泡板测试分级系统。

如果您有任何疑问，请随时在评论部分发表评论！

***但在您之前，请务必在下面的表单中输入您的电子邮件地址，以便在PyImageSearch博客上发布未来的教程时收到通知！***

## 下载： ##

![download-code-icon.png][] 如果您想下载此帖子中使用的代码和图片，请在下面的表单中输入您的电子邮件地址。您不仅会获得.zip的代码，我还会向您发送一份 **17页** 的 **免费计算机视觉，OpenCV和深度学习资源指南。**在里面你会找到我精心挑选的教程，书籍，课程和图书馆，以帮助你掌握简历和DL！听起来不错？如果是这样，请输入您的电子邮件地址，我会立即将代码发送给您！

电子邮件地址：

Website

window.rmpanda = window.rmpanda || \{\};  
window.rmpanda.wpvars = window.rmpanda.wpvars || \{\};

function pipelistToArray(pipelist) \{  
return (pipelist || ‘’).replace(/^|||$/g, ‘’).split(’|’);  
\}

function arrayToPipelist(array) \{  
return ‘|’ + (array || \[\]).join(’|’) + ‘|’;  
\}

function createLeaderboard(cachedPosts, dataKey) \{  
var leaderboard = \{\};  
var totalRead = cachedPosts.length;

cachedPosts.forEach(function(pageview, idx) {
    	var pID = pageview[0]; var slugs = pageview[1][dataKey];
    	var score = 0.5 + ((1.0 - (idx / totalRead)) * 0.5);
    	slugs.forEach(function(slug) {
    		leaderboard[slug] = leaderboard[slug] || 0;
    		leaderboard[slug] += score;
    	});
    });
    
    return leaderboard;

window.rmpanda.wpvars = \{\};

var cachedPosts = JSON.parse(localStorage.getItem(‘rmCachedPosts’)) || \[\];

var categoriesLeaderboard = createLeaderboard(cachedPosts, ‘categories’);  
logger.log(“CATEGORIES LEADERBOARD:”, categoriesLeaderboard);

var tagsLeaderboard = createLeaderboard(cachedPosts, ‘tags’);  
logger.log(“TAGS LEADERBOARD:”, tagsLeaderboard);

window.rmpanda.wpvars.favoriteCategorySlugs = arrayToPipelist(findLeaderboardWinners(categoriesLeaderboard));  
window.rmpanda.wpvars.favoriteTagSlugs = arrayToPipelist(findLeaderboardWinners(tagsLeaderboard));

localStorage.setItem(‘rmCachedPosts’, JSON.stringify(cachedPosts));

### 资源指南（完全免费）。 ###

在下面输入您的电子邮件地址，获取我的 **免费17页计算机视觉，OpenCV和深度学习资源指南PDF** 。在里面你会找到我精心挑选的教程，书籍，课程和Python库，以帮助你掌握计算机视觉和深度学习！

<form action="https://www.getdrip.com/forms/657075648/submissions" method="post" target="_blank" data-drip-embedded-form="657075648" data-drip-id="657075648">
    			<div>
    				<input type="email" name="fields[email]" class="required email" value="" placeholder="Your email address" style="width: 50%;">
    			</div>
    			<div>
    				<font style="vertical-align: inherit;"><font style="vertical-align: inherit;"><input id="mc-embedded-subscribe" type="submit" name="submit" class="btn submit button" value="下载指南！" data-drip-attribute="sign-up-button"></font></font>
    			</div>
    		<input type="hidden" name="time_zone" value="Asia/Shanghai"><input type="hidden" name="url" value="https://www.pyimagesearch.com/2016/10/03/bubble-sheet-multiple-choice-scanner-and-test-grader-using-omr-python-and-opencv/"><input type="hidden" name="page_title" value="Bubble sheet multiple choice scanner and test grader using OMR, Python and OpenCV - PyImageSearch"></form>
    		
    					<div class="social">
    	   					</div>
    		
    	</div>
    
    	
        <div class="fix"></div>
    </aside>
    <div class="post-utility"><p class="tags"><i class="icon-tag"></i> <a href="https://www.pyimagesearch.com/tag/bubble-sheet/" rel="tag"><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">气泡板</font></font></a><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">，</font></font><a href="https://www.pyimagesearch.com/tag/exam/" rel="tag"><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">考试</font></font></a><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">，</font></font><a href="https://www.pyimagesearch.com/tag/omr/" rel="tag"><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">omr</font></font></a><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">，</font></font><a href="https://www.pyimagesearch.com/tag/optical-mark-recognitionb/" rel="tag"><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">光学标记识别b</font></font></a><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">，</font></font><a href="https://www.pyimagesearch.com/tag/scantron/" rel="tag"><font style="vertical-align: inherit;"><font style="vertical-align: inherit;">scantron</font></font></a></p> </div>

[Adrian Rosebrock]: https://www.pyimagesearch.com/author/adrian/
[Link 1]: https://www.pyimagesearch.com/category/image-processing/
[Link 2]: https://www.pyimagesearch.com/category/tutorials/
[14]: /images/20211023/8118bcdb68654d46b81c65d7dc1cb576.png
[Link 3]: https://www.pyimagesearch.com/2014/09/01/build-kick-ass-mobile-document-scanner-just-5-minutes/
[Link 4]: https://www.pyimagesearch.com/2015/04/20/sorting-contours-using-python-and-opencv/
[Link 5]: https://www.pyimagesearch.com/2014/08/25/4-point-opencv-getperspective-transform-example/
[1]: /images/20211023/dd7ea99e13ab4240b239d07aa2baf2df.png
[imutils]: https://github.com/jrosebr1/imutils
[2]: /images/20211023/449fc0cb677b4db0a18bcd74667b0b54.png
[3]: /images/20211023/7a7da7b27f884dea9f5c21b4eb86ebc0.png
[Link 6]: https://www.pyimagesearch.com/2016/03/21/ordering-coordinates-clockwise-with-python-and-opencv/
[4]: /images/20211023/bfafbf4a9e2f43798b61e0b9c7c89b3d.png
[5_Otsu]: /images/20211023/ce8b5c29738942a39c9f0569fcfaf6e4.png
[6]: /images/20211023/cf457c89754c4e16bda54c0017526901.png
[7_]: /images/20211023/b7eebefdcfe04fa192ad9cb59ec4df03.png
[8]: /images/20211023/a3e22ec1f8a04573af4e4aef19022aaa.png
[9]: /images/20211023/074eb47ddfe54756bd4c0401b7ab142a.png
[10_OMR]: /images/20211023/b4806b973c294720a0b97c0e3f646daf.png
[Link 7]: https://www.pyimagesearch.com/2014/07/21/detecting-circles-images-using-opencv-hough-circles/
[11]: /images/20211023/6801e8ee44fa4688823908c671922a5a.png
[12_Python_OpenCV]: /images/20211023/4aca63c76e534b9bbdd446c3c8803cd8.png
[13_Python_OpenCV]: /images/20211023/3d0f659ba3a0438bb36375bcc5f2ffdb.png
[OMR]: https://en.wikipedia.org/wiki/Timeboxing
[15]: /images/20211023/9d18bd4db2484ae08caa0088bfd1490f.png
[16]: /images/20211023/929bb8b44b3443188ab22a5dad8dd8e6.png
[download-code-icon.png]: /images/20211023/6face9c26fe1429099432a1b1619c838.png