Systems, methods, and devices relating to computer vision-basedcounting of individuals in a crowd. A first video frame is usedto define a region of interest. The number of people in theregion of interest is estimated. A second or subsequent videoframe is then Used to adjust the boundaries of the region ofinterest. This is done by tracking the pixels of the boundaryof the region of interest. The number of people entering andleaving the region of interest between the first and secondvideo frames can then be determined by estimating the number ofpeople within either a union or an intersection of the originalregion of interest and the adjusted region of interest.
展开▼
