Shape Template Based Side-View Car Detection Algorithm

Abstract

In this paper, we describe a side-view car detection algorithm based on template detection. The template is constructed according to car shape knowledge. The algorithm can roughly be divided into two stages, hypothesis generation and hypothesis verification. In HG step, we use Hough transform to detect wheels, with the shape knowledge we get a rectangular area to generate hypothesis area. In HV step, least squares based fitting line detection algorithm is applied to pick out the line segments in the hypothesis area of edge image, and then we detect cars by the contour knowledge. By the above process the car is detected in the image. This paper also describes the least squares fitting line detection algorithm. The experimental results of many car images show that our method has fine accuracy in car detection and costs less time.