Study Of The Graphic/Image Preprocessing Technology In The Vehicle Type Automatic Recognition System

In these recent years, along with the development of our country's economy, traffic transportation becomes more and more busy, and it is a new request for traffic management. In practical transport system, such as highway and bridge toll station, city road monitor system, port and airdrome, police col system, large-scale parking lot, so as to emphasize the importance of vehicle type automatic recognition, to charge the payment for it, and to enhance automatic degree of vehicle monitoring and management. Therefore, it is very importance to carry out modern management of all toll station.As for vehicle type automatic recognition technology based on neural network, this paper mainly does some in-depth research from four parts as follows:First, as for vehicle type automatic recognition system, this paper discusses deeply in the study background and meaning of it, analyzes particularly the application and research of it, indicates the objection and lack of it, and shows the each part of it. This paper also indicates some theory, image file format, hardware, software, simulation of recognition system.Second, this paper discusses deeply in several main factors of system of basing on neural network's vehicle type automatic recognition, and proposes a couple of new methods. Firstly we analyze vehicle image collected, by means of image pre-process, which to improve quality of image. Secondly, image segmentation using threshold selection technique, etc., and then extract features from it. Then we use a trained off-line BP (Back Propagation) neural network to recognize vehicle type. Results proved all the algorithms proposed are effective to whether static or motion vehicles.Third, we design arithmetic based on Sobel operator for vehicle type automatic recognition system. We eliminate background images with serial Image Difference Algorithm and eliminate discrete noise. We adopt horizontal/vertical filling to outline the vehicle image. This processing method has simple and high speed recognition rate. In addition, we discuss the Canny edge detection operator, it can reserve the vehicle edge feature and have a good result for low-quality images.Finally, we establish a neural network's vehicle type automatic recognition system, to check the feasibility of our theory research, and improve the arithmetic by the continuing practical test, we achieve a pleased result.The result shows that the preprocessing method we mentioned is very simple and has a quick recognition speed, high recognition rate to the vehicle side image. It settles a good fundamental technology for later computer pattern recognition, and also improves the pattern recognition's precision and processing rate.