Real-time vision-based vehicle detection for rear-end collision mitigation systems

Abstract

This paper describes a real-time vision-based system that detects vehicles approaching from the rear in order to anticipate possible rear-end collisions. A camera mounted on the rear of the vehicle provides images which are analysed by means of computer vision techniques. The detection of candidates is carried out using the top-hat transform in combination with intensity and edge-based symmetries. The candidates are classified by using a Support Vector Machine-based classifier (SVM) with Histograms of Oriented Gradients (HOG features). Finally, the position of each vehicle is tracked using a Kalman filter and template matching techniques. The proposed system is tested using image data collected in real traffic conditions. © 2009 Springer-Verlag Berlin Heidelberg.