find_feature_matches(img1,img2,keypoints_1,keypoints_2,matches)函数

find_feature_matches

vector<KeyPoint> keypoints_1,keypoints_2;
vector<DMatch> matches;
find_feature_matches(img1,img2,keypoints_1,keypoints_2,matches);

1. 定义

void find_feature_matches ( const Mat& img1, const Mat& img2,
    						vector<KeyPoint>& kp1,
	  						vector<KeyPoint>& kp2,
	    					vector<DMatch>& matches)
{
  Mat dp1,dp2;
  Ptr<FeatureDetector> detector = ORB::create();
  Ptr<DescriptorExtractor> descriptor = ORB::create();
  Ptr<DescriptorMatcher> matcher = DescriptorMatcher::create("BruteForce-Hamming");

  detector->detect (img1, kp1);
  detector->detect (img2, kp2);

  descriptor->compute( img1, kp1, dp1);
  descriptor->compute( img2, kp2, dp2);

  vector<DMatch> match;
  matcher->match( dp1, dp2, match);

  double min_d = 10000,max_d = 0;

  for (int i = 0; i < dp1.rows; ++i) {
    double dist = match[i].distance;
	if(dist < min_d) min_d = dist;
	if(dist > max_d) max_d = dist;
  }

  printf("max dist : %f \n",max_d);
  printf("min dist : %f \n",min_d);

  for (int i =0; i < dp1.rows;++i) {
    if (match[i].distance <= max( 2*min_d, 30.0 ))
	  matches.push_back(match[i]);
  }

}

2. KeyPoint类

//默认构造函数
CV_WRAP KeyPoint():pt(0,0),size(0),angle(-1),response(0),octave(0),class_id(-1){}

//pt(x,y):关键点坐标
//size():该关键点邻域直径大小
//angle:角度，表示关键点的方向，值为【0,360】，负值表示不使用
//response:响应强度
//octave:从哪一层金字塔得到的关键点
//class_id：当要对图片进行分类时，用class_id对每个关键点进行区分，默认为-1

关于keypoints_1[0]，keypoints_2[0]，matches[]：
keypoints_1[0]：存储图一的第0个特征点，不一定与keypoints_2[0]对应
keypoints_2[0]：存储图二的第0个特征点，不一定与keypoints_2[0]对应
matches[0]:存储某一特征点的信息，图一中第matches.queryIdx个特征点与图二中第matches.trainIdx个特征点是相互对应的，matches.distance是两个对应特征点之间的距离。