FREAK描述子的使用，基于opencv-3.2.0

如果按照上述title搜索，网上的demo都来自于论文的算法实现：github-freak，但是最新的3.2.0的接口是完全不适用的，果然是contrib…
opencv官方代码见链接

需要做以下修正

• 头文件和命名空间（新的代码结构）

    #include <opencv2/features2d/features2d.hpp>    
    #include <opencv2/xfeatures2d.hpp>
    #include <opencv2/xfeatures2d/nonfree.hpp>
    
    using namespace cv::xfeatures2d; 
  

• 如果使用 FREAK extractor; 会出现error : The function/feature is not implemented () in detectAndCompute，应该使用下面的写法

  Ptr<FREAK> extractor = FREAK::create();
  

  详细见链接：stackoverflow；官方文档也有说明：官方链接

  BFMatcher matcher; 也是类似的情况…如果是 BruteForceMatcher<HammingLUT> matcher_Freak; 的写法要更新…关于 BFMatcher 类型还需要注意： 链接
  

• 关于 selectPairs 函数

  2.*有说明；3.0的文档只会说明 create 函数。

  官方链接

  Select the 512 best description pair indexes from an input (grayscale) image set. FREAK is available with a set of pairs learned offline. Researchers can run a training process to learn their own set of pair.

  FREAK描述子因为会互相覆盖，所以本身包含的信息就会存在冗余。论文中有说会通过排序只选择前512个pair！FREAK算法本身会提供一套训练好的pairs供使用…selectPairs 函数的作用就是可以让开发者自己重新训练一套pairs…

  stackoverflow也有类似的解答：链接

• 剔除outliers

  可以使用RANSAC进行剔除：见 链接

• 最终的实现

      #include <iostream>
      #include <string>
      #include <vector>
      
      #include <opencv2/core/core.hpp>
      #include <opencv2/highgui/highgui.hpp>
      // 下面三个头文件是FREAK需要的，路径和2.*完全不同
      #include <opencv2/features2d/features2d.hpp>    
      #include <opencv2/xfeatures2d.hpp>
      #include <opencv2/xfeatures2d/nonfree.hpp>
      
      using namespace cv;
      using namespace cv::xfeatures2d;        // 此命名空间也是3.0之后新加的
      using namespace std;
      
      void help( char** argv )
      {
          std::cout << "\nUsage: " << argv[0] << " [path/to/image1] [path/to/image2] \n"
                    << "This is an example on how to use the keypoint descriptor presented in the following paper: \n"
                    << "A. Alahi, R. Ortiz, and P. Vandergheynst. FREAK: Fast Retina Keypoint. \n"
                    << "In IEEE Conference on Computer Vision and Pattern Recognition, 2012. CVPR 2012 Open Source Award winner \n"
                    << std::endl;
      }
      
      int main( int argc, char** argv ) {
          // check http://opencv.itseez.com/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.html
          // for OpenCV general detection/matching framework details
      
          if( argc != 3 ) {
              help(argv);
              return -1;
          }
      
          // Load images
          Mat imgA = imread(argv[1], CV_LOAD_IMAGE_GRAYSCALE );
          if( !imgA.data ) {
              std::cout<< " --(!) Error reading image " << argv[1] << std::endl;
              return -1;
          }
      
          Mat imgB = imread(argv[2], CV_LOAD_IMAGE_GRAYSCALE );
          if( !imgA.data ) {
              std::cout << " --(!) Error reading image " << argv[2] << std::endl;
              return -1;
          }
      
          std::vector<KeyPoint> keypointsA, keypointsB;
          Mat descriptorsA, descriptorsB;
          std::vector<DMatch> matches;
      
          // DETECTION
          // Any openCV detector such as
          Ptr<cv::xfeatures2d::SurfFeatureDetector> detector;
          detector = cv::xfeatures2d::SurfFeatureDetector::create(2000);
      
          // DESCRIPTOR
          Ptr<FREAK> extractor = FREAK::create();
      
          // MATCHER
          // The standard Hamming distance can be used such as
          BFMatcher matcher(NORM_HAMMING);        // 使用Hamming距离测试
          
          // detect
          double t = (double)getTickCount();
          detector->detect( imgA, keypointsA );
          detector->detect( imgB, keypointsB );
          t = ((double)getTickCount() - t)/getTickFrequency();
          std::cout << "detection time [s]: " << t/1.0 << std::endl;
      
          // extract
          t = (double)getTickCount();
          extractor->compute( imgA, keypointsA, descriptorsA );
          extractor->compute( imgB, keypointsB, descriptorsB );
          t = ((double)getTickCount() - t)/getTickFrequency();
          std::cout << "extraction time [s]: " << t << std::endl;
      
          // match
          t = (double)getTickCount();
          matcher.match(descriptorsA, descriptorsB, matches);
          t = ((double)getTickCount() - t)/getTickFrequency();
          std::cout << "matching time [s]: " << t << std::endl;
      
          // Draw matches
          Mat imgMatch;
          drawMatches(imgA, keypointsA, imgB, keypointsB, matches, imgMatch); // 系统函数
      
          namedWindow("matches", CV_WINDOW_KEEPRATIO);
          imshow("matches", imgMatch);
          waitKey(0);
      }