Vehicle size estimation from stereoscopic video

The scope of the paper is stereoscopic CCTV with applications to road traffic surveillance. The main idea is to exploit stereoscopic video analysis for automatic estimation of dimensions of moving vehicles. The major steps of the proposed technique include stereoscopic video correction, identification of characteristic points of vehicle contours in video frames, depth estimation and vehicle physical length calculation. The technique has been tested using an experimental system described in the paper. The paper also comprises experimental results obtained in real environment for cars moving along a city street. The obtained results prove sufficient accuracy of the technique.