Research Of Image Analysis Algorithm In Video Vehicle Detection

Transportation is very important in national economy and society development. Vehicle detection is an important part of transportation management system. With image processing technology developing, Video vehicle detection is widely used in vehicle velocity detection. This article studies image analysis algorithm in video vehicle detection.The algorithm is composed of four parts: image prior process, background reconstruction, object segment and match, coordinate transform.Prior process the images which are obtained by camera fixed above road. The way is that change color image into gray image and filter the prime image with medial filter.This paper uses longest-stable-state algorithm to reconstruct background. The prime way of this algorithm is that given in long-time sequence frames, the gray value of background pixel is in longest-stable-state in these sequence frames. The gray value in longest-stable-state is the pixel gray value in background. So it can reconstruct background.This paper puts up a moving object segmentation algorithm from image sequence based on edge. Because the vehicle gray maybe similar to the road and influence of shadow, direct difference will bring wrong object segment. This algorithm is proposed by combining edge detection, Gauss filter, and background difference. The algorithm reinforces the edge of moving object by edge detection, and avoids the influence of noise brought by edge detection. Finally, the moving object is segmented by background difference. After object segment, this paper uses region character tracking to tract object in sequence images based on geometry character. So it can obtain situation difference of the object.According to the situation difference of the object, this paper uses a rectification method for perspective distortion image to transform coordinate. So it can change pixel situation difference into actual situation difference, and then calculate the vehicle velocity.