A Low-Cost Occlusion Handling Using a Novel Feature in Congested Traffic Images

Occlusion detection is critical for robust and reliable vision-based systems for traffic flow analysis. In this paper we discuss novel occlusion detection approach. The goal of this algorithm is to detect vehicles in congested traffic scenes with occlusion handling. Also we proposed a new proper feature for vehicle detection to avoid the merging of two or more objects into one and to improve the accuracy of object localization. This feature is strong shadow which exists under any vehicle. Many sequences of real images including noisy images, cluttered, and images containing remarkable shadows and occlusions, taken from both scenes of outdoor highways and city roads, have been taken for evaluating our method