Background suppression for video vehicle tracking systems with moving cameras using camera motion estimation

Abstract

Camera oscillations and slight movements are typical in the video based parts of the Intelligent Transportation Systems, especially in the cases when the cameras are mounted on the high pylons or pillars, similarly as some street lamps. The influence of strong wind and some vibrations caused by some heavy vehicles may result in some shifts of the images captured as the consecutive video frames. In such situations some typical background estimation and removal algorithms based on the comparison of corresponding pixels of each video frame may lead to significant errors. The influence of such camera motions increases seriously for high focal length corresponding to tracking of distant objects. In order to minimize the influence of such movements the background suppression algorithm using the camera motion estimation is proposed in the paper increasing the stability of the estimated background which is further used in the vehicle tracking algorithm. © 2012 Springer-Verlag.