Use SVM to solve the problem of Level 3 Classification of 2D Spatial Vectors

[Original: blog. csdn. netfirefightarticledetails6400060] To Learn OPENCVSVM classifier, refer to the use of SVM on the Internet to solve the problem of binary vector classification and change it to C code, for reference only: OPENCV2.2VS2008 step: 1, generate random points and classify them according to a certain spatial distribution. 2,

[Original article: http://blog.csdn.net/firefight/article/details/6400060] to learn the opencv svm classifier, use SVM on the Internet to solve the classification problem of 2D Spatial vectors and change it to C code. For reference only: Step 1, generate random points and classify them according to a certain spatial distribution. 2,

[Original article: http://blog.csdn.net/firefight/article/details/6400060]

To learn the opencv svm classifier, refer to "Using SVM to solve the classification problem of 2D Spatial vectors" on the Internet and change it to C ++ code for reference only.

Environment: OPENCV2.2 + VS2008

Steps:
1. generate random points and classify them according to a certain spatial distribution.
2. Create SVM and use random vertex samples for training
3. Divide the entire space by SVM classification results and display the Support Vector

[Cpp]View plaincopy

1. # Include "stdafx. h"
2. # Include
4. Void drawCross (Mat & img, Point center, Scalar color)
5. {
6. Int col = center. x> 2? Center. x: 2;
7. Int row = center. y> 2? Center. y: 2;
9. Line (img, Point (col-2, row-2), Point (col + 2, row + 2), color );
10. Line (img, Point (col + 2, row-2), Point (col-2, row + 2), color );
11. }
13. Int newSvmTest (int rows, int cols, int testCount)
14. {
15. If (testCount> rows * cols)
16. Return 0;
18. Mat img = Mat: zeros (rows, cols, CV_8UC3 );
19. Mat testPoint = Mat: zeros (rows, cols, CV_8UC1 );
20. Mat data = Mat: zeros (testCount, 2, CV_32FC1 );
21. Mat res = Mat: zeros (testCount, 1, CV_32SC1 );
23. // Create random test points
24. For (int I = 0; I <testCount; I ++)
25. {
26. Int row = rand () % rows;
27. Int col = rand () % cols;
29. If (testPoint. (Row, col) = 0)
30. {
31. TestPoint. (Row, col) = 1;
32. Data. (I, 0) = float (col)/cols;
33. Data. (I, 1) = float (row)/rows;
34. }
35. Else
36. {
37. I --;
38. Continue;
39. }
41. If (row> (50 * cos (col * CV_PI/100) + 200 ))
42. {
43. DrawCross (img, Point (col, row), CV_RGB (255, 0, 0 ));
44. Res. (I, 0) = 1;
45. }
46. Else
47. {
48. If (col> 200)
49. {
50. DrawCross (img, Point (col, row), CV_RGB (0,255, 0 ));
51. Res. (I, 0) = 2;
52. }
53. Else
54. {
55. DrawCross (img, Point (col, row), CV_RGB (0, 0,255 ));
56. Res. (I, 0) = 3;
57. }
58. }
60. }
62. // Show test points
63. Imshow ("dst", img );
64. WaitKey (0 );
66. //// // Start svm trainning //////////////////
67. CvSVM svm = CvSVM ();
68. CvSVMParams param;
69. CvTermCriteria criteria;
71. Criteria = cvTermCriteria (CV_TERMCRIT_EPS, 1000, FLT_EPSILON );
72. /* SVM type: CvSVM: C_SVC
    Types of Kernel: CvSVM: RBF
    Degree: 10.0 (not used this time)
    Gamma: 8.0
    Coef0: 1.0 (not used this time)
    C: 10.0
    Nu: 0.5 (not used this time)
    P: 0.1 (not used this time)
    Then the training data is normalized and placed in an array of the CvMat type. */
    
73. Param = CvSVMParams (CvSVM: C_SVC, CvSVM: RBF, 10.0, 8.0, 1.0, 10.0, 0.5, NULL, criteria );
74. Svm. train (data, res, Mat (), Mat (), param );
76. For (int I = 0; I <rows; I ++)
77. {
78. For (int j = 0; j <cols; j ++)
79. {
80. Mat m = Mat: zeros (1, 2, CV_32FC1 );
81. M. (0, 0) = float (j)/cols;
82. M. (0, 1) = float (I)/rows;
84. Float ret = 0.0;
85. Ret = svm. predict (m );
86. Scalar rcolor;
88. Switch (int) ret)
89. {
90. Case 1: rcolor = CV_RGB (100, 0, 0); break;
91. Case 2: rcolor = CV_RGB (0,100, 0); break;
92. Case 3: rcolor = CV_RGB (0, 0,100); break;
93. }
95. Line (img, Point (j, I), Point (j, I), rcolor );
96. }
97. }
99. Imshow ("dst", img );
100. WaitKey (0 );
102. // Show support vectors
103. Int sv_num = svm. get_support_vector_count ();
104. For (int I = 0; I <sv_num; I ++)
105. {
106. Const float * support = svm. get_support_vector (I );
107. Circle (img, Point (int) (support [0] * cols), (int) (support [1] * rows), 5, CV_RGB (200,200,200 ));
108. }
110. Imshow ("dst", img );
111. WaitKey (0 );
113. Return 0;
114. }
116. Int main (int argc, char ** argv)
117. {
118. Return newSvmTest (400,600,100 );
119. }

Learning samples:

Category:

Support Vector: