A Vehicle Detection Method via Symmetry in Multi-Scale Windows

Abstract

A novel method based upon multi-scale windows for efficient detection of the vehicles is described in this paper. This method presumes that the local image of the vehicle is the most symmetrical on the center of vehicle while the least symmetrical on the edge of vehicle. However, the symmetrical value of the center of the vehicle would not be the only maximum value in the actual images, which causes troubles in locating vehicles. We present a new method based on multi-scale windows in which both the symmetry axis and the edges of the vehicle can be detected simultaneously. We also concocted a strategy of sequential processing in order to improve the detection. Experimental results shows that our algorithm performances well in detecting vehicles in front of our mobile.