Vehicle Color Extraction Based on First Sight Window

Currently, most research in vehicle color extraction is on a complete or static vehicle image or under a strict restriction on the position of vehicles and makes it hardly be applied in real-time applications. Firstly, new concepts such as First Sight Window and Second Sight Regions are proposed in this paper. Secondly, several algorithms are developed based on the first sight window and bottom-up hierarchical search strategy to extract vehicle color from the sequence of traffic images taken by some fixed cameras or sensors over city roads. Experimental results show that our methods are efficient with high and correct rate of recognition and little management time delay, so it can be applied in real time.