作者：Taily老段
opencv查找轮廓---cvFindContours && cvDrawCountours 用法及例子
1、找到轮廓；
2、每个连通域的均值中心；
3、求连通域半径（平均）；
4、相似度——最小半径/最大半径；
5、根据相似度阈值、半径阈值来判断是否是圆
1. #include <iostream>
2. #include <opencv2/imgproc/imgproc.hpp>
3. #include <opencv2/core/core.hpp>
4. #include <opencv2/highgui/highgui.hpp>
5. using namespace cv;
6. using namespace std;
7.
8.
9. int main()
10. {
11. Mat q_MatImage;
12. Mat q_MatImageGray;
13. Mat q_MatImageShow;
14. Mat q_MatImageShow2;
15. q_MatImage=imread("1.png");//读入一张图片
16. q_MatImage.copyTo(q_MatImageShow);
17. q_MatImage.copyTo(q_MatImageShow2);
18. cvtColor(q_MatImage,q_MatImageGray,CV_RGB2GRAY);
19. double q_dEpsilon = 10E-9;
20. unsigned int q_iReturn=0;
21.
22. int q_iX,q_iY,q_iWidth,q_iHeight;
23. q_iX=20;
24. q_iY=40;
25. q_iWidth=600;
26. q_iHeight=420;
27.
28. double q_dThresholdSimilarity=60;
29. double q_dThresholdMin=35;
30. double q_dThresholdMax=75;
31.
32. // Rect q_RectROI = Rect(q_iX,q_iY,q_iWidth,q_iHeight);
33. // Mat q_MatROI = q_MatImageGray(q_RectROI);
34. //
35. // threshold(q_MatROI, q_MatROI, 150, 255, CV_THRESH_BINARY);
36. threshold(q_MatImageGray, q_MatImageGray, 150, 255, CV_THRESH_BINARY);
37.
38. namedWindow("Test1"); //创建一个名为Test窗口
39. imshow("Test1",q_MatImageGray); //窗口中显示图像
40.
41. vector<vector<Point>> q_vPointContours;
42.
43. //findContours(q_MatROI, q_vPointContours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_NONE,Point(q_iX,q_iY));
44. findContours(q_MatImageGray, q_vPointContours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_NONE,Point(0,0));
45.
46. size_t q_iAmountContours = q_vPointContours.size();
47.
48. for (size_t i = 0;i < q_iAmountContours; i++)
49. {
50. size_t q_perNum = q_vPointContours[i].size();
51. for (size_t j = 0;j < q_iAmountContours; j++)
52. {
53. circle( q_MatImageGray, q_vPointContours[i][j] ,3 , CV_RGB(0,255,0),1, 8, 3 );
54. }
55. }
56.
57. namedWindow("findContours");
58. imshow("findContours",q_MatImageGray);
59.
60. std::vector<cv::Point2f> q_vPointCentersContours(q_iAmountContours);
61. std::vector<double> q_vdRadiusesContours(q_iAmountContours);
62. std::vector<double> q_vdSimilarityContours(q_iAmountContours);
63. std::vector<bool> q_vbFlagCircles(q_iAmountContours);
64.
65. std::vector<double> q_vdRadiusesContour;
66. double q_dRadiusMax,q_dRadiusMin;
67. double q_dSumX,q_dSumY;
68. size_t q_iAmountPoints;
69.
70. for(size_t q_iCycleContours=0;q_iCycleContours<q_iAmountContours;q_iCycleContours++)
71. {
72. q_dSumX=0.0;
73. q_dSumY=0.0;
74. q_iAmountPoints=q_vPointContours[q_iCycleContours].size();
75. if(0>=q_iAmountPoints)
76. {
77. continue;
78. }
79. for(size_t q_iCyclePoints=0;q_iCyclePoints<q_iAmountPoints;q_iCyclePoints++)
80. {
81. q_dSumX+=q_vPointContours[q_iCycleContours].at(q_iCyclePoints).x;
82. q_dSumY+=q_vPointContours[q_iCycleContours].at(q_iCyclePoints).y;
83. }
84.
85. q_vPointCentersContours[q_iCycleContours].x=(float)(q_dSumX/q_iAmountPoints);//均值中心点X</span>
86. q_vPointCentersContours[q_iCycleContours].y=(float)(q_dSumY/q_iAmountPoints);//均值中心点Y</span>
87.
88.
89. q_vdRadiusesContour.resize(q_iAmountPoints);
90. double q_dDifferenceX,q_dDifferenceY;
91. double q_dSumRadius=0.0;
92. q_dRadiusMax=0.0;
93. q_dRadiusMin=DBL_MAX;;
94. for(size_t q_iCyclePoints=0;q_iCyclePoints<q_iAmountPoints;q_iCyclePoints++)
95. {
96. q_dDifferenceX=q_vPointCentersContours[q_iCycleContours].x-q_vPointContours[q_iCycleContours].at(q_iCyclePoints).x;
97. q_dDifferenceY=q_vPointCentersContours[q_iCycleContours].y-q_vPointContours[q_iCycleContours].at(q_iCyclePoints).y;
98. q_vdRadiusesContour[q_iCyclePoints]=sqrt(q_dDifferenceX*q_dDifferenceX+q_dDifferenceY*q_dDifferenceY);
99.
100. if(q_vdRadiusesContour[q_iCyclePoints]>q_dRadiusMax)
101. {
102. q_dRadiusMax=q_vdRadiusesContour[q_iCyclePoints];
103. }
104. if(q_vdRadiusesContour[q_iCyclePoints]<q_dRadiusMin)
105. {
106. q_dRadiusMin=q_vdRadiusesContour[q_iCyclePoints];
107. }
108.
109. q_dSumRadius+=q_vdRadiusesContour[q_iCyclePoints];
110. }
111. q_vdRadiusesContours[q_iCycleContours]=q_dSumRadius/q_iAmountPoints; //均值半径
112.
113. q_vdSimilarityContours[q_iCycleContours]=100.0*q_dRadiusMin/q_dRadiusMax; //相似度
114. if((q_dThresholdSimilarity<q_vdSimilarityContours[q_iCycleContours])&&
115. (q_dThresholdMin<q_vdRadiusesContours[q_iCycleContours])&&
116. (q_dThresholdMax>q_vdRadiusesContours[q_iCycleContours])) //判断是否是圆
117. {
118. q_vbFlagCircles[q_iCycleContours]=true;
119. }
120. else
121. {
122. q_vbFlagCircles[q_iCycleContours]=false;
123. }
124. }
125.
126.
127. if(q_dEpsilon < 10)
128. {
129. cv::Point q_PointCenterCurrent;
130. for(size_t q_iCycleContours=0;q_iCycleContours<q_iAmountContours;q_iCycleContours++)
131. {
132. if(q_vbFlagCircles[q_iCycleContours])
133. {
134. q_PointCenterCurrent.x=(int)(q_vPointCentersContours[q_iCycleContours].x);
135. q_PointCenterCurrent.y=(int)(q_vPointCentersContours[q_iCycleContours].y);
136. circle(q_MatImageShow,q_PointCenterCurrent,3,Scalar(0.0,0.0,255.0),0);
137. }
138. }
139. }
140.
141. int q_iIndexResultBegin=4;
142. int q_iAmountCircleResult=4;
143. int q_iIndexCiecleCurrent;
144.
145. int q_iCountCircles=0;
146.
147. for(size_t q_iCycleContours=0;q_iCycleContours<q_iAmountContours;q_iCycleContours++)
148. {
149. if(q_vbFlagCircles[q_iCycleContours])
150. {
151. q_iIndexCiecleCurrent=q_iIndexResultBegin+q_iAmountCircleResult*q_iCountCircles;
152. // match_result[q_iIndexCiecleCurrent]=(float)(q_vdSimilarityContours[q_iCycleContours]);
153. // match_result[q_iIndexCiecleCurrent+1]=(float)(q_vdRadiusesContours[q_iCycleContours]);
154. // match_result[q_iIndexCiecleCurrent+2]=(float)(q_vPointCentersContours[q_iCycleContours].x);
155. // match_result[q_iIndexCiecleCurrent+3]=(float)(q_vPointCentersContours[q_iCycleContours].y);
156. q_iCountCircles++;
157. }
158. }
159. cout << "总共找到 " << q_iCountCircles << "个圆！" << endl;
160.
161.
162. namedWindow("Test"); //创建一个名为Test窗口
163. imshow("Test",q_MatImageShow);//窗口中显示图像
164. waitKey(); //等待5000ms后窗口自动关闭
165. } 
大律算法otsu：
1. int thresh = Otsu(q_MatImageGray);
2. threshold(q_MatImageGray, q_MatImageGray, thresh, 255, CV_THRESH_BINARY);
3.
4. for(int i=0; i < q_MatImageGray.rows; i++)
5. {
6. for(int j = 0; j < q_MatImageGray.cols; j++)
7. {
8. q_MatImageGray.at<uchar>(i,j) = 255 -q_MatImageGray.at<uchar>(i,j);
9. }
10. } 1. int Otsu(Mat src)
2. {
3. int height=src.rows;
4. int width =src.cols;
5.
6.
7. //histogram
8. float histogram[256] = {0};
9. for(int i=0; i < height; i++)
10. {
11. unsigned char* p=(unsigned char*)src.ptr<uchar>(i);
12. for(int j = 0; j < width; j++)
13. {
14. histogram[*p++]++;
15. }
16. }
17. //normalize histogram
18. int size = height * width;
19. for(int i = 0; i < 256; i++)
20. {
21. histogram[i] = histogram[i] / size;
22. }
23.
24.
25. //average pixel value
26. float avgValue=0;
27. for(int i=0; i < 256; i++)
28. {
29. avgValue += i * histogram[i]; //整幅图像的平均灰度
30. }
31.
32.
33. int threshold;
34. float maxVariance=0;
35. float w = 0, u = 0;
36. for(int i = 0; i < 256; i++)
37. {
38. w += histogram[i]; //假设当前灰度i为阈值, 0~i 灰度的像素(假设像素值在此范围的像素叫做前景像素) 所占整幅图像的比例
39. u += i * histogram[i]; // 灰度i 之前的像素(0~i)的平均灰度值： 前景像素的平均灰度值
40.
41.
42. float t = avgValue * w - u;
43. float variance = t * t / (w * (1 - w) );
44. if(variance > maxVariance)
45. {
46. maxVariance = variance;
47. threshold = i;
48. }
49. }
50.
51.
52. return threshold;
53. } 1. int Otsu2(Mat src)
2. {
3. int height=src.rows;
4. int width =src.cols;
5.
6. int x=0,y=0;
7. int pixelCount[256];
8. float pixelPro[256];
9. int i, j, pixelSum = width * height, threshold = 0;
10.
11. //初始化
12. for(i = 0; i < 256; i++)
13. {
14. pixelCount[i] = 0;
15. pixelPro[i] = 0;
16. }
17.
18. //统计灰度级中每个像素在整幅图像中的个数
19. for(i = y; i < height; i++)
20. {
21. for(j = x;j <width;j++)
22. {
23. pixelCount[src.at<uchar>(i,j)]++;
24. }
25. }
26.
27.
28. //计算每个像素在整幅图像中的比例
29. for(i = 0; i < 256; i++)
30. {
31. pixelPro[i] = (float)(pixelCount[i]) / (float)(pixelSum);
32. }
33.
34. //经典ostu算法,得到前景和背景的分割
35. //遍历灰度级[0,255],计算出方差最大的灰度值,为最佳阈值
36. float w0, w1, u0tmp, u1tmp, u0, u1, u,deltaTmp, deltaMax = 0;
37. for(i = 0; i < 256; i++)
38. {
39. w0 = w1 = u0tmp = u1tmp = u0 = u1 = u = deltaTmp = 0;
40.
41. for(j = 0; j < 256; j++)
42. {
43. if(j <= i) //背景部分
44. {
45. //以i为阈值分类，第一类总的概率
46. w0 += pixelPro[j];
47. u0tmp += j * pixelPro[j];
48. }
49. else //前景部分
50. {
51. //以i为阈值分类，第二类总的概率
52. w1 += pixelPro[j];
53. u1tmp += j * pixelPro[j];
54. }
55. }
56.
57. u0 = u0tmp / w0; //第一类的平均灰度
58. u1 = u1tmp / w1; //第二类的平均灰度
59. u = u0tmp + u1tmp; //整幅图像的平均灰度
60. //计算类间方差
61. deltaTmp = w0 * (u0 - u)*(u0 - u) + w1 * (u1 - u)*(u1 - u);
62. //找出最大类间方差以及对应的阈值
63. if(deltaTmp > deltaMax)
64. {
65. deltaMax = deltaTmp;
66. threshold = i;
67. }
68. }
69. //返回最佳阈值;
70. return threshold;
71. } 关注【OpenCV学习交流】