静态背景中的运动目标检测识别跟踪

#include "cxcore.h" #include "cv.h" #include "highgui.h" # include <iostream> #include <math.h> # define NULL 0 const int winHeight = 480; const int winWidth = 640; static int nFrmNum = 0; struct objectList { int x; int y; struct objectList *next; }; struct movingObject { int area; int label; int x; int y; CvPoint points[2]; //矩形框的两个对脚点 struct objectList *head; struct movingObject *next; }; void bin_img_or(IplImage *image, IplImage *image1, IplImage *image2); void bg_diff(IplImage *pFrame, IplImage *pBkImg, IplImage *pFrImg, CvMat *pFrameMat, CvMat *pBkMat, CvMat *pFrMat); void computeObject(CvMat* foreImageMat, int width, int height, CvMat* labeledMat, movingObject* head); int cal_dist(int x1, int y1, int x2, int y2); void release_obj_link(movingObject *link); void draw_line(IplImage *image, objectList *t); static int label_num = 0; int main() { int tmp_x, tmp_label, dist; int cover_flag = 0; CvMat* pFrameMat = NULL; CvMat* pFrMat = NULL; CvMat* pBkMat = NULL; IplImage *res = 0; //帧差视频 IplImage *res0 = 0; IplImage *image = 0; //原始帧 IplImage *image_pass = 0; IplImage *pBkImg = 0; //背景图像 IplImage *pFrImg = 0; //背景差分视频 IplImage *pFrame = 0; CvCapture *capture = NULL; CvMat* LabeledMat = NULL; movingObject *p_obj = NULL; const CvMat* prediction = NULL; CvPoint predict_pt; CvFont font1, font2; cvInitFont(&font1, CV_FONT_HERSHEY_TRIPLEX, 1, 1); cvInitFont(&font2, CV_FONT_HERSHEY_COMPLEX, 0.5, 0.5); /* CvFont font; cvInitFont(&font,CV_FONT_HERSHEY_SIMPLEX ,1,1);*/ capture = cvCaptureFromFile("D:/picture/cd2.mp4"); IplImage *frame = 0; frame = cvQueryFrame(capture); int image_width = (int)frame->width; //得到图像的宽度和高度 int image_height = (int)frame->height; image = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3); //创建图像首地址并分配存储空间，3通道 image_pass = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3); res = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1); //灰度图像 res0 = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3); pBkImg = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1); pFrImg = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1); pFrame = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1); //图像对应的矩阵都是一维的 pBkMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1); pFrMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1); pFrameMat = cvCreateMat(pFrame->height, pFrame->width, CV_32FC1); LabeledMat = cvCreateMat(image_height, image_width, CV_32FC1); movingObject *curr_head = new movingObject(); movingObject *prev_head = new movingObject(); cvCopy(frame, image_pass, NULL); while (1) { image = cvQueryFrame(capture); if (image == NULL) { break; } cvShowImage("原始", image); cvSmooth(image, image, CV_GAUSSIAN, 3, 0, 0, 0); cvAbsDiff(image, image_pass, res0); //计算两帧图像差的绝对值 cvCvtColor(res0, res, CV_RGB2GRAY); //转换为灰度图像 //cvThreshold(res, res, 50, 255, CV_THRESH_BINARY); cvAdaptiveThreshold(res, res, 255, CV_ADAPTIVE_THRESH_MEAN_C, CV_THRESH_BINARY_INV, 5, 5); //二值化 cvShowImage("帧差视频", res); cvShowImage("背景图像", pBkImg); bg_diff(image, pBkImg, pFrImg, pFrameMat, pBkMat, pFrMat); cvShowImage("背景差分视频", pFrImg); cvZero(pFrame); bin_img_or(pFrame, res, pFrImg); cvDilate(pFrame, pFrame, 0, 1); cvConvert(pFrame, pFrMat); cvCopy(image, image_pass, NULL); cvShowImage("或", pFrame); if (nFrmNum > 1) { computeObject(pFrMat, image_width, image_height, LabeledMat, curr_head); p_obj = curr_head->next; while (p_obj != NULL) { cvRectangle(image, cvPoint(p_obj->x - 1, p_obj->y - 1), cvPoint(p_obj->x + 1, p_obj->y + 1), cvScalar(0, 0, 255), 4, 8, 0); //画出目标的质心 p_obj = p_obj->next; } if (nFrmNum == 3) { computeObject(pFrMat, image_width, image_height, LabeledMat, prev_head); } if (nFrmNum > 3) { movingObject *q = NULL, *p = NULL; q = curr_head->next; while (q) { for (p = prev_head->next; p; p = p->next) { dist = cal_dist(p->x, p->y, q->x, q->y); if (dist <= 25) { break; } } if (p == NULL) { //生成新标签 q->label = ++label_num; } q = q->next; } for (q = curr_head->next; q; q = q->next) { for (p = prev_head->next; p; p = p->next) { dist = cal_dist(p->x, p->y, q->x, q->y); printf("dist is : %d\n", dist); if (dist <= 15) { q->label = p->label; q->head->next = p->head; p->head = NULL; } } draw_line(image, q->head); cvRectangle(image, q->points[0], q->points[1], cvScalar(0, 255, 0)); char buf[10]; sprintf_s(buf, 10, "%d", q->label); cvPutText(image, buf, cvPoint(q->points[0].x, q->points[0].y + 20), &font1, CV_RGB(255, 255, 0)); } printf("\n"); release_obj_link(prev_head->next); prev_head->next = curr_head->next; curr_head->next = NULL; } } char FrmNum[30]; sprintf_s(FrmNum, 30, "Frame No: %d", nFrmNum); cvPutText(image, FrmNum, cvPoint(1, 25), &font2, CV_RGB(0, 0, 255)); cvShowImage("video", image); int inter = cvWaitKey(16); if (inter == 27) { break; } if (inter == 32) { if (cvWaitKey(1500000) == 32) { } } } cvReleaseImage(&res); cvReleaseImage(&res0); cvReleaseImage(&image_pass); cvReleaseCapture(&capture); return 0; } void draw_line(IplImage *image, objectList *t) { objectList *t1 = t; while (t1 != NULL) //描述目标运动的轨迹 { if (t1->next != NULL) { cvLine(image, cvPoint(t1->x, t1->y), cvPoint(t1->next->x, t1->next->y), cvScalar(255, 0, 255), 2, 8, 0); } t1 = t1->next; } } void release_obj_link(movingObject *link) { movingObject *p = link, *q = NULL; objectList *t1 = NULL, *t2 = NULL; while (p != NULL) { q = p; p = p->next; t1 = q->head; q->head = NULL; while (t1 != NULL) { t2 = t1; t1 = t1->next; delete t2; } delete q; } } int cal_dist(int x1, int y1, int x2, int y2) { int distance; double res; res = (x1 - x2) * (x1 - x2) + (y1 - y2) * (y1 - y2); distance = (int)(sqrt(res)); return distance; } void bin_img_or(IplImage *image, IplImage *image1, IplImage *image2) { int i, j; int width = image->width; int height = image->height; unsigned char data1, data2, data; for (i = 0; i < height; i++) for (j = 0; j < width; j++) { data1 = (unsigned char)image1->imageData[i * width + j]; data2 = (unsigned char)image2->imageData[i * width + j]; if (data1 == 255 || data2 == 255) { data = 255; image->imageData[i * width + j] = (char)data; } } } void bg_diff(IplImage *pFrame, IplImage *pBkImg, IplImage *pFrImg, CvMat *pFrameMat, CvMat *pBkMat, CvMat *pFrMat) { nFrmNum++; if (nFrmNum == 1) { //将彩色图像转化为灰度图像 cvCvtColor(pFrame, pBkImg, CV_BGR2GRAY); cvCvtColor(pFrame, pFrImg, CV_BGR2GRAY); //将灰度图像存入矩阵，灰度图像是单通道的 cvConvert(pFrImg, pFrameMat); cvConvert(pFrImg, pFrMat); cvConvert(pBkImg, pBkMat); } else { cvCvtColor(pFrame, pFrImg, CV_BGR2GRAY); cvConvert(pFrImg, pFrameMat); //高斯滤波 cvSmooth(pFrameMat, pFrameMat, CV_GAUSSIAN, 3, 0, 0, 0); //当前帧减去背景图像并取绝对值 cvAbsDiff(pFrameMat, pBkMat, pFrMat); //二值化前景图像 cvThreshold(pFrMat, pFrImg, 30, 255.0, CV_THRESH_BINARY); //滑动平均更新背景（求平均） cvRunningAvg(pFrameMat, pBkMat, 0.01, 0); //将背景矩阵转化为图像格式，用以显示 cvConvert(pBkMat, pBkImg); } }