Contribution to the realization of a video surveillance system by optical flow and stereovision

In this paper we combine two techniques in order to detect, track and identify person actions in a stream video. The combination of optical flow and stereovision allows also reconstructing the 3D shape of a scene from its 2D images. We take images from two cameras and match points to determine disparity, and therefore depth. The optical flow method is used to estimate motion field in a video scene. Experimental results with synthetic and real sequences, exhibit the usefulness of our proposed approach.