OpenCV SITF 特征提取 FeatureDetector对象函数detect运行报错解决方案

本人在使用OpenCV SIFT特征提取算法时，遇到了问题，具体表现为 .exe触发了一个断点错误，经网上查找，发现是 vector 在析构时，造成了内存错误，解决方案由大神在csdn博客中给出，链接地址：

点击打开链接

按照大神所提示的 ，在使用opencv 函数之前，先对 vector 进行手动分配内存，分配内存后问题解决；

个人声明：本人在VS2013下调用sift算法，已经将问题解决，当然，具体问题，具体分析，本人适用，但不一定适用所有人，也希望大家对该问题的解决方案进行及时补充；

 

代码：

1. initModule_nonfree();//初始化模块，使用SIFT或SURF时用到

2.  

3. //FeatureDetector::create()函数的参数可选

4. // • "FAST" – FastFeatureDetector

5. // • "STAR" – StarFeatureDetector

6. // • "SIFT" – SIFT(nonfree module)

7. // • "SURF" – SURF(nonfree module)

8. // • "ORB" – ORB

9. // • "BRISK" – BRISK

10. // • "MSER" – MSER

11. // • "GFTT" – GoodFeaturesToTrackDetector

12. // • "HARRIS" – GoodFeaturesToTrackDetector with Harris detector enabled

13. // • "Dense" – DenseFeatureDetector

14. // • "SimpleBlob" – SimpleBlobDetector

15. Ptr<FeatureDetector> detector = FeatureDetector::create("SIFT");//创建SIFT特征检测器

16.  

17. //SiftFeatureDetector detector;

18. //Ptr<FeatureDetector> detector = FeatureDetector::create("SURF");

19. // DescriptorExtractor::create()函数的参数如下：

20. //

21. // • "SIFT" – SIFT

22. // • "SURF" – SURF

23. // • "BRIEF" – BriefDescriptorExtractor

24. // • "BRISK" – BRISK

25. // • "ORB" – ORB

26. // • "FREAK" – FREAK

27. //SiftDescriptorExtractor descriptor_extractor;

28. Ptr<DescriptorExtractor> descriptor_extractor = DescriptorExtractor::create("SIFT");//创建特征向量生成器

29. //Ptr<DescriptorExtractor> descriptor_extractor = DescriptorExtractor::create("SURF");

30. // // DescriptorMatcher::create()函数的参数如下，对应不同的匹配算法：

31. // // – BruteForce(it uses L2)

32. // // – BruteForce - L1

33. // // – BruteForce - Hamming

34. // // – BruteForce - Hamming(2)

35. // // – FlannBased

36. Ptr<DescriptorMatcher> descriptor_matcher = DescriptorMatcher::create("BruteForce");//创建特征匹配器,暴力匹配(简单匹配)

37. // if (detector.empty() || descriptor_extractor.empty())

38. // cout << "fail to create detector!";

39.  

40. //读入图像

41. Mat img1 = imread("001.jpg", 0);

42. Mat img2 = imread("004.bmp", 0);

43.  

44. if (img1.empty() || img1.channels() != 1)

45. MessageBox(L"图像类型错误");

46. //CV_Error(Error::StsBadArg, "image is empty or has incorrect depth (!=CV_8U)");

47.  

48. //if (!mask.empty() && mask.type() != CV_8UC1)

49. //CV_Error(Error::StsBadArg, "mask has incorrect type (!=CV_8UC1)");

50.  

51. //特征点检测

52. double t = getTickCount();//当前滴答数

53. vector<KeyPoint> keypoints1, keypoints2;

54. keypoints1.resize(100);

55. keypoints2.resize(100);

56. detector->detect(img1, keypoints1);//检测img1中的SIFT特征点，存储到keypoints1中

57. detector->detect(img2, keypoints2);

58. //detector.detect(img1, keypoints1);

59. //detector.detect(img2, keypoints2);

60. // cout << "图像1特征点个数:" << keypoints1.size() << endl;

61. // cout << "图像2特征点个数:" << keypoints2.size() << endl;

62. //

63. // //根据特征点计算特征描述子矩阵，即特征向量矩阵

64. Mat descriptors1, descriptors2;

65. descriptor_extractor->compute(img1, keypoints1, descriptors1);

66. descriptor_extractor->compute(img2, keypoints2, descriptors2);

67. // t = ((double)getTickCount() - t) / getTickFrequency();

68. // cout << "SIFT算法用时：" << t << "秒" << endl;

69. //

70. //

71. // cout << "图像1特征描述矩阵大小：" << descriptors1.size()

72. // << "，特征向量个数：" << descriptors1.rows << "，维数：" << descriptors1.cols << endl;

73. // cout << "图像2特征描述矩阵大小：" << descriptors2.size()

74. // << "，特征向量个数：" << descriptors2.rows << "，维数：" << descriptors2.cols << endl;

75. //

76. // //画出特征点

77. Mat img_keypoints1, img_keypoints2;

78. drawKeypoints(img1, keypoints1, img_keypoints1, Scalar::all(-1), 0);

79. drawKeypoints(img2, keypoints2, img_keypoints2, Scalar::all(-1), 0);

80. imshow("Src1",img_keypoints1);

81. imshow("Src2",img_keypoints2);

82. waitKey(0);

83.  

84. //特征匹配

85. vector<DMatch> matches;//匹配结果

86. matches.resize(100);

87. descriptor_matcher->match(descriptors1, descriptors2, matches);//匹配两个图像的特征矩阵

88. cout << "Match个数：" << matches.size() << endl;

89.  

90. //计算匹配结果中距离的最大和最小值

91. //距离是指两个特征向量间的欧式距离，表明两个特征的差异，值越小表明两个特征点越接近

92. double max_dist = 0;

93. double min_dist = 100;

94. for (int i = 0; i < matches.size(); i++)

95. {

96. double dist = matches[i].distance;

97. if (dist < min_dist) min_dist = dist;

98. if (dist > max_dist) max_dist = dist;

99. }

100. cout << "最大距离：" << max_dist << endl;

101. cout << "最小距离：" << min_dist << endl;

102.  

103. //筛选出较好的匹配点

104. vector<DMatch> goodMatches;

105. for (int i = 0; i < matches.size(); i++)

106. {

107. if (matches[i].distance < 0.5 * max_dist)

108. {

109. goodMatches.push_back(matches[i]);

110. }

111. }

112. cout << "goodMatch个数：" << goodMatches.size() << endl;

113.  

114. //画出匹配结果

115. Mat img_matches;

116. //红色连接的是匹配的特征点对，绿色是未匹配的特征点

117. drawMatches(img1, keypoints1, img2, keypoints2, goodMatches, img_matches,

118. /*Scalar::all(-1)/ **/CV_RGB(255,0,0), CV_RGB(0, 255, 0), Mat(), 2);

119.  

120. namedWindow("MatchSIFT", 0);

121. imshow("MatchSIFT", img_matches);

122. waitKey(0);