OpenCV中feature2D学习——BFMatcher和FlannBasedMatcher

ゞ浴缸里的玫瑰 2022-08-13 17:47 26阅读 0赞

# Brute Force匹配和FLANN匹配是opencv二维特征点匹配常见的两种办法，分别对应[BFMatcher][]（BruteForceMatcher）和[FlannBasedMatcher][]。BFMatcher的构造函数如下： **C++:** BFMatcher::BFMatcher(int **normType**=NORM\_L2, bool **crossCheck**=false ) <table style="font-size:16px; margin-bottom:10px; border:0px; border-collapse:collapse; margin-top:20px"> <tbody> <tr> <th style="padding:1px 8px 1px 5px; border:0px!important">Parameters:</th> <td style="padding:1px 8px 1px 5px; border:0px!important"> <ul style="margin:0px; padding-left:1em"> <li style="text-align:justify; line-height:20px; margin:0px">normType&nbsp;– One of&nbsp;NORM_L1,&nbsp;NORM_L2,&nbsp;NORM_HAMMING,&nbsp;NORM_HAMMING2.&nbsp;L1&nbsp;and&nbsp;L2&nbsp;norms are preferable choices for SIFT and SURF descriptors,&nbsp;NORM_HAMMING&nbsp;should be used with ORB, BRISK and BRIEF,&nbsp;NORM_HAMMING2&nbsp;should be used with ORB when&nbsp;WTA_K==3&nbsp;or&nbsp;4&nbsp;(see ORB::ORB constructor description).</li> <li style="text-align:justify; line-height:20px; margin:0px">crossCheck&nbsp;– If it is false, this is will be default BFMatcher behaviour when it finds the k nearest neighbors for each query descriptor. If&nbsp;crossCheck==true, then the&nbsp;knnMatch()&nbsp;method with&nbsp;k=1&nbsp;will only return pairs&nbsp;(i,j)&nbsp;such that for&nbsp;i-th&nbsp;query descriptor the&nbsp;j-th&nbsp;descriptor in the matcher’s collection is the nearest and vice versa, i.e. the&nbsp;BFMatcher&nbsp;will only return consistent pairs. Such technique usually produces best results with minimal number of outliers when there are enough matches. This is alternative to the ratio test, used by D. Lowe in SIFT paper.</li> </ul> </td> </tr> </tbody> </table> FlannBasedMatcher的构造函数如下： **class** **FlannBasedMatcher** : **public** DescriptorMatcher \{ **public**: FlannBasedMatcher( **const** Ptr<flann::IndexParams>& indexParams=**new** flann::KDTreeIndexParams(), **const** Ptr<flann::SearchParams>& searchParams=**new** flann::SearchParams() ); **virtual** void add( **const** vector<Mat>& descriptors ); **virtual** void clear(); **virtual** void train(); **virtual** bool isMaskSupported() **const**; **virtual** Ptr<DescriptorMatcher> clone( bool emptyTrainData=**false** ) **const**; **protected**: ... \}; 二者的区别在于BFMatcher总是尝试所有可能的匹配，从而使得它总能够找到最佳匹配，这也是Brute Force（暴力法）的原始含义。而FlannBasedMatcher中FLANN的含义是Fast Library forApproximate Nearest Neighbors，从字面意思可知它是一种近似法，算法更快但是找到的是最近邻近似匹配，所以当我们需要找到一个相对好的匹配但是不需要最佳匹配的时候往往使用FlannBasedMatcher。当然也可以通过调整FlannBasedMatcher的参数来提高匹配的精度或者提高算法速度，但是相应地算法速度或者算法精度会受到影响。 此外， 使用特征提取过程得到的特征描述符（descriptor）数据类型有的是float类型的，比如说SurfDescriptorExtractor， SiftDescriptorExtractor，有的是uchar类型的，比如说有ORB，BriefDescriptorExtractor。对应float类型的匹配方式有：FlannBasedMatcher，BruteForce<L2<float>>，BruteForce<SL2<float>>，BruteForce<L1<float>>。对应uchar类型的匹配方式有：BruteForce<Hammin>，BruteForce<HammingLUT>。所以ORB和BRIEF特征描述子只能使用BruteForce匹配法。 #

[BFMatcher]: http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html#bfmatcher
[FlannBasedMatcher]: http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html#flannbasedmatcher