Research On Technology Of Vehicle Anti-collision Warning System Based On Vision

With the change of times and the rapid development of science and technology, cars are becoming more and more integrated into the rhythm of people’s life. However, traffic security problems what caused by the car also become more and more serious and can not be ignored. In view of this problem, how to avoid the traffic accidents caused by other vehicles and obstacles on the road has become one of the hot spots in today’s research. Comparing with ultrasonic, laser, radar, the technology based on vision is cheaper and more suitable for the human eye’s information capture habits. Therefore, the technology of vehicle anti-collision warning system based on vision is an effective way to solve the above problems.Through analyzing and comparing the various algorithms, the corresponding vehicle detection and tracking algorithms based on the domestic and foreign intelligent vehicle systems are designed. The main research contents are as follows:(1) Through the study of vanishing point, obtain the region which to be detected; then in order to highlight the characteristics of the lane line, use the Sobel edge detection algorithm to enhance the road region. On the basis of this, adopt the improved Hough transform to detect the lane line, and finally obtain the ROI region.(2) In the ROI region, put forward the OTSU algorithm combined with GA to segment the target vehicle and road background and then use the shadow feature of vehicle bottom to extract the front target vehicle to get the potential target location.(3) Combined with the existing vehicle characteristics and contour information, obtain the vertical and horizontal projection histograms. On the basic of the characteristics of the vehicle itself with the vertical edge symmetry, verify the authenticity of the target and exclude false vehicles by judging whether the ROI is a vehicle region.(4) Through the study of vehicle distance measurement, the conversion relation between different coordinate systems is established. And the internal and external parameters of the camera are also demarcated. The distance formula is derived based on geometric projection relation with the camera calibration method. Then the parameters are substituted into the formula to measure the distance between the car and the front car. Finally the system interface is constructed and the warning strategy is formulated.Experimental results show that the proposed methods can detect and track the target vehicle in front, and also can estimate the distance from the vehicle and the vehicle ahead, which provides a powerful guarantee for the traffic security problems.