利用OpenCV检测手掌(palm)和拳头(fist)

思路：利用训练好的palm.xml和fist.xml文件，用OpenCV的CascadeClassifier对每一帧图像检测palm和fist，之后对多帧中检测到的palm和fist进行聚类分组，满足分组条件的区域为最终检测结果。

代码：

 #include "opencv2/objdetect/objdetect.hpp" #include "opencv2/highgui/highgui.hpp" #include "opencv2/imgproc/imgproc.hpp" #include <iostream> #include <stdio.h> using namespace std; using namespace cv; /** Function Headers */ void detectAndDisplay( Mat frame ); void RestoreVectors(vector<vector<Rect>>& vecs_bank, vector<Rect>& vecAll); /** Global variables */ String palm_cascade_name = "palm.xml"; String fist_cascade_name = "fist.xml"; CascadeClassifier palm_cascade; CascadeClassifier fist_cascade; string window_name = "Capture - Palm and fist detection"; /** @function main */ int main( int argc, const char** argv ) {   CvCapture* capture;   Mat frame;   //-- 1. Load the cascades   if( !palm_cascade.load( palm_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };   if( !fist_cascade.load( fist_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };   //-- 2. Read the video stream   capture = cvCaptureFromCAM( -1 );   if( capture )   {     while( true )     {   frame = cvQueryFrame( capture );   //-- 3. Apply the classifier to the frame       if( !frame.empty() )       { detectAndDisplay( frame ); }       else       { printf(" --(!) No captured frame -- Break!"); break; }       int c = waitKey(10);       if( (char)c == 'q' || (char)c == 'Q' || 27 == c) { break; }      }   }   cvReleaseCapture(&capture);   return 0; }/** @function detectAndDisplay */void detectAndDisplay( Mat frame ){  std::vector<Rect> faces;  std::vector<Rect> palms;  std::vector<Rect> fists;  static vector<vector<Rect>> palms_bank;  static vector<vector<Rect>> fists_bank;  const int MAX_NUM = 3;  Mat frame_gray;  cvtColor( frame, frame_gray, CV_BGR2GRAY );  equalizeHist( frame_gray, frame_gray );  //-- Palm detection  palm_cascade.detectMultiScale( frame_gray, palms, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );  palms_bank.push_back(palms);  if(palms_bank.size() > MAX_NUM)  palms_bank.erase(palms_bank.begin());  vector<Rect> palmAll;  RestoreVectors(palms_bank, palmAll);  groupRectangles(palmAll, 2);   for( size_t j = 0; j < palmAll.size(); j++ )  {rectangle(frame, palmAll[j], Scalar(0,255,0), 2);  }  //-- Fist detection  fist_cascade.detectMultiScale( frame_gray, fists, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );  fists_bank.push_back(fists);  if(fists_bank.size() > MAX_NUM)  fists_bank.erase(fists_bank.begin());  vector<Rect> fistAll;  RestoreVectors(fists_bank, fistAll);  groupRectangles(fistAll, 2);   for( size_t j = 0; j < fistAll.size(); j++ )  {rectangle(frame, fistAll[j], Scalar(0,0,255), 2);  }  //-- Show what you got  imshow( window_name, frame ); }void RestoreVectors(vector<vector<Rect>>& vecs_bank, vector<Rect>& vecAll){for(size_t i = 0; i < vecs_bank.size(); i++){vecAll.insert(vecAll.end(), vecs_bank[i].begin(), vecs_bank[i].end());}}

首先，分类器palm_cascade和fist_cascade分别读入palm.xml和fist.xml文件；

然后，分类器palm_cascade和fist_cascade分别调用detectMultiScale函数对输入的灰度图像进行检测，检测的结果是一系列Rect区域，分别存入palms和fists中；

之后，将每帧图像检测的结果palms和fists再存入palms_bank和fists_bank中，palms_bank和fists_bank中保存了MAX_NUM帧的检测结果；

之后，调用RestoreVectors函数，将palms_bank和fists_bank中的结果重新存入vector结构中(因为groupRectangles接受的参数是vector<Rect>，而不是vector<vector<Rect>>)，并调用groupRectangles进行聚类。

最后，将聚类后的结果画出来，palm用绿色画出，fist用红色画出。

groupRectangles的说明如下：

groupRectangles对rectList中的Rect进行聚类，近似大小和近似位置的Rect被分为一类(cluster)，只有当一类中的Rect数目超过groupThreshold时，该类别才会被保留，仍保留在rectList中。

参考：

[1] groupRectangles的说明文档

[2] palm.xml和fist.xml的下载地址

[3] 人脸和眼睛检测的opencv示例代码

代码，palm.xml和fist.xml文件，说明文档可以从这里下载：

http://download.csdn.net/detail/lichengyu/7751671

groupRectangles

Groups the object candidate rectangles.

C++: void groupRectangles(vector<Rect>& rectList, int groupThreshold, double eps=0.2)

C++: void groupRectangles(vector<Rect>& rectList, vector<int>& weights, intgroupThreshold, double eps=0.2)

Python: cv2.groupRectangles(rectList, groupThreshold[, eps]) → rectList, weights

Parameters:

• rectList – Input/output vector of rectangles. Output vector includes retained and grouped rectangles. (The Python list is not modified in place.)
• groupThreshold – Minimum possible number of rectangles minus 1. The threshold is used in a group of rectangles to retain it.
• eps – Relative difference between sides of the rectangles to merge them into a group.

The function is a wrapper for the generic function partition() . It clusters all the input rectangles using the rectangle equivalence criteria that combines rectangles with similar sizes and similar locations. The similarity is defined by eps. When eps=0 , no clustering is done at all. If  , all the rectangles are put in one cluster. Then, the small clusters containing less than or equal to groupThreshold rectangles are rejected. In each other cluster, the average rectangle is computed and put into the output rectangle list.