opencv3/C++ FLANN特征匹配方式

使用函数detectAndCompute(）检测关键点并计算描述符

函数detectAndCompute(）参数说明：

1. void detectAndCompute(
2. InputArray image, //图像
3. InputArray mask, //掩模
4. CV_OUT std::vector<KeyPoint>& keypoints,//输出关键点的集合
5. OutputArray descriptors,//计算描述符（descriptors[i]是为keypoints[i]的计算描述符）
6. bool useProvidedKeypoints=false //使用提供的关键点
7. );

match(）从查询集中查找每个描述符的最佳匹配。

参数说明：

1. void match(
2. InputArray queryDescriptors, //查询描述符集
3. InputArray trainDescriptors, //训练描述符集合
4. CV_OUT std::vector<DMatch>& matches, //匹配
5. InputArray mask=noArray() //指定输入查询和描述符的列表矩阵之间的允许匹配的掩码
6. ) const;

FLANN特征匹配示例：

1. #include<opencv2/opencv.hpp>
2. #include<opencv2/xfeatures2d.hpp>
3. using namespace cv;
4. using namespace cv::xfeatures2d;
5.  
6. //FLANN对高维数据较快
7. int main()
8. {
9. Mat src1,src2;
10. src1 = imread("E:/image/image/card2.jpg");
11. src2 = imread("E:/image/image/cards.jpg");
12. if (src1.empty() || src2.empty())
13. {
14. printf("can ont load images....\n");
15. return -1;
16. }
17. imshow("image1", src1);
18. imshow("image2", src2);
19.  
20. int minHessian = 400;
21. //选择SURF特征
22. Ptr<SURF>detector = SURF::create(minHessian);
23. std::vector<KeyPoint>keypoints1;
24. std::vector<KeyPoint>keypoints2;
25. Mat descriptor1, descriptor2;
26. //检测关键点并计算描述符
27. detector->detectAndCompute(src1, Mat(), keypoints1, descriptor1);
28. detector->detectAndCompute(src2, Mat(), keypoints2, descriptor2);
29.  
30. //基于Flann的描述符匹配器
31. FlannBasedMatcher matcher;
32. std::vector<DMatch>matches;
33. //从查询集中查找每个描述符的最佳匹配
34. matcher.match(descriptor1, descriptor2, matches);
35. double minDist = 1000;
36. double maxDist = 0;
37. for (int i = 0; i < descriptor1.rows; i++)
38. {
39. double dist = matches[i].distance;
40. printf("%f \n", dist);
41. if (dist > maxDist)
42. {
43. maxDist = dist;
44. }
45. if (dist < minDist)
46. {
47. minDist = dist;
48. }
49.  
50. }
51. //DMatch类用于匹配关键点描述符的
52. std::vector<DMatch>goodMatches;
53. for (int i = 0; i < descriptor1.rows; i++)
54. {
55. double dist = matches[i].distance;
56. if (dist < max(2.5*minDist, 0.02))
57. {
58. goodMatches.push_back(matches[i]);
59. }
60. }
61. Mat matchesImg;
62. drawMatches(src1, keypoints1, src2, keypoints2, goodMatches, matchesImg, Scalar::all(-1), Scalar::all(-1), std::vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS);
63. imshow("output", matchesImg);
64.  
65. waitKey();
66. return 0;
67. }

以上这篇opencv3/C++ FLANN特征匹配方式就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持我们。

原文链接：https://blog.csdn.net/akadiao/article/details/79163727