视频前景目标提取.pdf-资料库

视频前景目标提取 1. 2. #include 3. #include 4. #include 5. #include 6. using namespace cv; 7. using namespace std; 8. 9. int main() 10. { 11. 12. //int argc, char* argv[] CvCapture *capture = cvCreateFileCapture("E:\\数学建模\\input2.avi "); 13. IplImage *img, *img_gray, *img_cny, *back, *fore;//原图像、灰度图像、边缘检测得到的图像、背景图像、前景图像 14. 15. 16. 17. 18. 19. 20. CvVideoWriter *writer=NULL; double fps = cvGetCaptureProperty(capture, CV_CAP_PROP_FPS); CvSize size = cvSize( (int)cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH), (int)cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_HEIGHT)); writer = cvCreateVideoWriter("E:\\数学建模\\input2out.avi", CV_FOU RCC('M','J','P','G'), fps, size, 0); 21. 22. 23. 24. 25. 26. 27. int width, height;//图像的长、宽 int cnt = 0; //视频帧数 img = cvQueryFrame(capture); cnt++; width = img->width; height = img->height; vector< vector > mean(width*height), sd(width*height), w(w idth*height); 28. 29. 30. 31. 32. 33. 34. 35. img_gray = cvCreateImage(cvSize(width, height), IPL_DEPTH_8U, 1); img_cny = cvCreateImage(cvSize(width, height), IPL_DEPTH_8U, 1); back = cvCreateImage(cvSize(width, height), IPL_DEPTH_8U, 1); fore = cvCreateImage(cvSize(width, height), IPL_DEPTH_8U, 1); cvCvtColor(img, img_gray, CV_BGR2GRAY); cvCopy(img_gray, back); //初始化混合高斯背景模型 int K = 5, sd_init = 16, T = 200, match = 0;//高斯分布的个数、标准差的初始值帧数阈值匹配参数

36. double w_init = 0.005, D = 2.5, alph, p, threold = 0.7;//权重的初始值、置信参数、学习速率、参数更新速率、预定阈值（用于选取背景模型） 37. 38. 39. 40. 41. for (int i = 0; i < height; i++) { for (int j = 0; j < width; j++) { mean[i*width + j].push_back((uchar)img_gray->imageData[i*widt h + j]); 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. sd[i*width + j].push_back(sd_init); w[i*width + j].push_back(1); } } IplImage* logpolar_frame = cvCreateImage(size, IPL_DEPTH_8U, 3); while (1) { img = cvQueryFrame(capture); if (!img) break; cvCvtColor(img, img_gray, CV_BGR2GRAY); cnt++; //计算学习速率 if (cnt < T) alph = (double)1 / (2 * cnt); else alph = (double)1 / (2 * T); for (int i = 0; i < height; i++) { for (int j = 0; j < width; j++) { match = 0; int len = mean[i*width + j].size(); for (int k = 0; k < len; k++) { //判断一个像素是否与背景模型匹配 if (fabs((uchar)img_gray->imageData[i*width + j] - mea n[i*width + j][k]) < D*sd[i*width + j][k]) 69. 70. 71. 72. alph; 73. 74. { match = 1; //更新权重、均值、标准差 w[i*width + j][k] = (1 - alph)*w[i*width + j][k] + p = alph / w[i*width + j][k]; mean[i*width + j][k] = (1 - p)*mean[i*width + j][k] + p*(uchar)img_gray->imageData[i*width + j];

75. sd[i*width + j][k] = sqrt((1 - p)*sd[i*width + j] [k] * sd[i*width + j][k] + p*pow((uchar)img_gray->imageData[i*width + j] - mean[i*width + j][k], 2)); w[i*width + j][k] = (1 - alph)*w[i*width + j][k]; 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. } else { } } if (match == 0) { //增加新的背景模型 if (len < K) { mean[i*width + j].push_back((uchar)img_gray->image Data[i*width + j]); 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. sd[i*width + j].push_back(sd_init); w[i*width + j].push_back(w_init); } //更改权重最小的背景模型 else { int min = 0; double tmp = w[i*width + j][0]; for (int k = 1; k < len; k++) { min = k; tmp = w[i*width + j][k]; if (tmp > w[i*width + j][k]) { } mean[i*width + j][min] = (uchar)img_gray->imag } eData[i*width + j]; 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. } sd[i*width + j][min] = sd_init; w[i*width + j][min] = w_init; } //每隔50 帧删除权重最小的高斯分布 if (cnt % 50 == 0) { vector::iterator it_w, it_sd, it_mean; it_w = w[i*width + j].begin(); it_sd = sd[i*width + j].begin();

115. 116. it_mean = mean[i*width + j].begin(); for (; it_w != w[i*width + j].end() && it_sd != sd [i*width + j].end() && it_mean != mean[i*width + j].end();) { if (*it_w < w_init && (*it_w / *it_sd) < (w_ini it_w = w[i*width + j].erase(it_w); it_sd = sd[i*width + j].erase(it_sd); it_mean = mean[i*width + j].erase(it_mean); 117. 118. t / sd_init)) 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. 151. 152. 153. 154. 155. 156. tmp += w[i*width + j][k]; } it_w++; it_sd++; it_mean++; { } else { } } //权重归一化 len = mean[i*width + j].size(); double tmp = 0; for (int k = 0; k < len; k++) { } for (int k = 0; k < len; k++) { } //计算各背景模型的优先级，并以从大到小的顺序排序 double *rand;//优先级 int *rand_ind;//排序后的优先级索引 rand = (double *)malloc(sizeof(double)*len); rand_ind = (int *)malloc(sizeof(int)*len); for (int k = 0; k < len; k++) { } for (int k = 1; k< len; k++) { for (int l = 0; l < k; l++) { w[i*width + j][k] /= tmp; rand[k] = w[i*width + j][k] / sd[i*width + j][k]; rand_ind[k] = k;

if (rand[k] > rand[l]) { double rand_tmp = rand[k]; rand[k] = rand[l]; rand[l] = rand_tmp; double rand_ind_tmp = rand_ind[k]; rand_ind[k] = rand_ind[l]; rand_ind[l] = rand_ind_tmp; 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. } } int temp = rand_ind[k]; sum += w[i*width + j][temp]; if (sum>threold) { } } double sum = 0; int B; for (int k = 0; k < len; k++) { } match = 0; back->imageData[i*width + j] = 0; //求背景模型 for (int k = 0; k <= B; k++) { B = k; break; int temp = rand_ind[k]; back->imageData[i*width + j] += w[i*width + j][tem p] * mean[i*width + j][temp]; 189. 190. 191. 192. 193. 194. 195. } /* //求前景图像 for (int k = 0; k <= B; k++) { int temp = rand_ind[k]; if (fabs((uchar)img_gray->imageData[i*width + j] - mea n[i*width + j][temp]) < D*sd[i*width + j][temp]) 196. 197. 198. { match = 1; break;

199. 200. 201. 202. } } if (match == 1) fore->imageData[i*width + j] = 0; else fore->imageData[i*width + j] = (uchar)img_gray->i mageData[i*width + j]; 203. 204. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. } } cvAbsDiff(img_gray, back, fore); cvCanny(img_gray, img_cny, 40, 100,3); cvShowImage("canny", img_cny); cvAnd(img_cny, fore, fore); cvDilate(fore, fore, NULL, 3); cvErode(fore, fore, NULL, 3); cvThreshold(fore, fore, 55, 255, CV_THRESH_BINARY); cvWriteFrame(writer, fore); */ free(rand); free(rand_ind); } cvDestroyAllWindows(); cvReleaseCapture(&capture); cvReleaseImage(&img_gray); cvReleaseImage(&fore); cvReleaseImage(&back); cvReleaseImage(&img_cny); cvReleaseVideoWriter(&writer); return (0); } cvShowImage("fore", fore); cvShowImage("back", back); cvShowImage("src", img); char s = cvWaitKey(1); if (s == 27) break;

资料库

视频前景目标提取.pdf

相关推荐

人工智能

热门标签

最新资料