Monocular based road vehicle detection with feature fusion and cascaded Adaboost algorithm

To decrease vehicle crash, a new rear view vehicle detection system based on monocular vision is designed. First, a small and flexible hardware platform based on DM642 DSP micro-controller is built. Then, a two-step vehicle detection algorithm is proposed. In the first step, a fast vehicle edge and symmetry fusion algorithm is obtained and a low threshold is set so that all possible vehicles will selected and also maintain a high false detection rate. In the second step, a classifier using Haar and HoG fusion feature based Adaboost classifier is trained to classify the possible vehicles and eliminate the false detected vehicles from the candidate vehicles generated in the first step. Experimental results demonstrate that this classifier maintains very high detection rate and low false detection rate in different road, weather and lighting conditions.