利用光流法计算人体运动的速度与方向

利用光流法计算人体运动的速度与方向

1.方向的计算 首先计算图像各个象素的光流(opencv LK),然后建立4*4窗口对X,Y方向分别做统计求和, 然后求得 atan(yy/xx)作为光流方向,即为运动的方向.

2.速度的计算 利用帧差分得到运动图像,然后建立4*4窗口对图像进行统计求和,求和值作为权重,表示速度的比例. 即运动区域白色(255)面积越大,速度越大.

3.结果 大部分运动方向计算正确,少部分有错误,还需要改进算法.(利用统计?)

4.代码:

WW_RETURN HumanMotion::ImgOpticalFlow(IplImage *pre_grey,IplImage *grey) /************************************************* Function: Description: 光流法计算运动速度与方向 Date: 2006－6－14 Author: Input: Output: Return: Others: *************************************************/ {

IplImage *velx = cvCreateImage( cvSize(grey->width ,grey->height),IPL_DEPTH_32F, 1 ); IplImage *vely = cvCreateImage( cvSize(grey->width ,grey->height),IPL_DEPTH_32F, 1 );

velx->origin = vely->origin = grey->origin; CvSize winSize = cvSize(5,5); cvCalcOpticalFlowLK( prev_grey, grey, winSize, velx, vely ); cvAbsDiff( grey,prev_grey, abs_img ); cvThreshold( abs_img, abs_img, 29, 255, CV_THRESH_BINARY);

CvScalar xc,yc; for(int y =0 ;y<velx->height; y++) for(int x =0;x<velx->width;x++ ) { xc = cvGetAt(velx,y,x); yc = cvGetAt(vely,y,x);

float x_shift= (float)xc.val[0]; float y_shift= (float)yc.val[0]; const int winsize=5; //计算光流的窗口大小

if((x%(winsize*2)==0) && (y%(winsize*2)==0) ) {

if(x_shift!=0 || y_shift!=0) { if(x>winsize && y>winsize && x <(velx->width-winsize) && y<(velx->height-winsize) ) {

cvSetImageROI( velx, cvRect( x-winsize, y-winsize, 2*winsize, 2*winsize)); CvScalar total_x = cvSum(velx); float xx = (float)total_x.val[0]; cvResetImageROI(velx);

cvSetImageROI( vely, cvRect( x-winsize, y-winsize, 2*winsize, 2*winsize)); CvScalar total_y = cvSum(vely); float yy = (float)total_y.val[0]; cvResetImageROI(vely); cvSetImageROI( abs_img, cvRect( x-winsize, y-winsize, 2*winsize, 2*winsize)); CvScalar total_speed = cvSum(abs_img); float ss = (float)total_speed.val[0]/(4*winsize*winsize)/255; cvResetImageROI(abs_img);

const double ZERO = 0.000001; const double pi = 3.1415926; double alpha_angle;

if(xx<ZERO && xx>-ZERO) alpha_angle = pi/2; else alpha_angle = abs(atan(yy/xx)); if(xx<0 && yy>0) alpha_angle = pi - alpha_angle ; if(xx<0 && yy<0) alpha_angle = pi + alpha_angle ; if(xx>0 && yy<0) alpha_angle = 2*pi - alpha_angle ;

CvScalar line_color; float scale_factor = ss*100; line_color = CV_RGB(255,0,0); CvPoint pt1,pt2; pt1.x = x; pt1.y = y; pt2.x = static_cast<int>(x + scale_factor*cos(alpha_angle)); pt2.y = static_cast<int>(y + scale_factor*sin(alpha_angle));

cvLine( image, pt1, pt2 , line_color, 1, CV_AA, 0 ); CvPoint p; p.x = (int) (pt2.x + 6 * cos(alpha_angle - pi / 4*3)); p.y = (int) (pt2.y + 6 * sin(alpha_angle - pi / 4*3)); cvLine( image, p, pt2, line_color, 1, CV_AA, 0 ); p.x = (int) (pt2.x + 6 * cos(alpha_angle + pi / 4*3)); p.y = (int) (pt2.y + 6 * sin(alpha_angle + pi / 4*3)); cvLine( image, p, pt2, line_color, 1, CV_AA, 0 );

/* line_color = CV_RGB(255,255,0); pt1.x = x-winsize; pt1.y = y-winsize; pt2.x = x+winsize; pt2.y = y+winsize; cvRectangle(image, pt1,pt2,line_color,1,CV_AA,0); */

} } } }

cvShowImage( "Contour", abs_img); cvShowImage( "Contour2", vely);

cvReleaseImage(&velx); cvReleaseImage(&vely); cvWaitKey(20); return WW_OK;

}