Preceding vehicle distance computation based on dark prior

This paper presents a monocular vision system for the detection and distance computation of preceding vehicles under foggy weather conditions. It is divided in three major steps. Firstly, adaboost algorithm is used to detect vehicles in front of the host vehicle. Secondly, these vehicles will be tracked by camshaft algorithm; the initial points for camshift algorithm are the centers of the preceding vehicles detected by adaboost algorithm. Thirdly the distance between these vehicles and the host vehicle will be calculated based on dark prior algorithm and monocular model. Finally, experimental results are given. 90% of preceding vehicles are detected correctly. Distance data for three preceding vehicles are listed in the form of graphic and table. The distance between these vehicles and the host vehicle calculated by our algorithm is close to the real distance observed by manual count. Experimental results verify the effectiveness of the algorithm.