The Research On Vehicle Objects Detection And Recognition Based On Video Image

Along with the development of digital image processing, video processing has been used in traffic detection step by step. Using one or more vidicons to collect the information of traffic status on driveway, we can obtain the information of vehicle type, speed, vehicle flux and so on by image processing. It can work with information management system to realize traffic control. Because of the advantage of easily fixing, good reliability and abundant visible information, video detection and processing system becomes the direction of the development of traffic control system and the research hot topic at home and abroad presently.In the process of video obtaining, outside environment variation such as little dithering of vidicon, slow variational daylight and the sway of tree will affect the presicion of moving target detection. Aiming at this problem, the paper presents a method of background reconstruction, in which there are moving targets in scene. The background can be dynamicly updated. It can decrease the influence of outside environment variation and shorten the time of measurement.In the process of video image processing, the shadows and inanitions of target will affect the prisicion of orientation. The paper locates the position of targets accurately by detecting target area, filtering noise, markering connected cell, filling inanitions, detecting target and eliminating shadows.In the process of vehicle recognition, the variation of geometric form will affect the recognition prisicion. Aiming at this problem, the paper presents two methods. The one is a method of vehicle recognition and classification based on characteristics of contour.Eigenvector has been first structured by using edge invariance, posture ratio, rectangle degree, elongation degree and roundness degree. Then the vehicle type recognition and classification is successfully realized by Euclidean distance. Another one is a method based on the orientation and matching of contour. Hausdorff distance has the advantage of processing in nosie and the variation of geometric form. So Hausdorff distance is used to recognize and classify the vehicle type. Because of more calculation of Hausdorff distance, the paper presents an improved method. It has been matched by using external rectangle in advance and then by Hausdorff distance accurately. It can shorten greatly the time of measurement. Radar speed measurement system has the problem of complicated equipment, poor stability, no visual information. Aiming at this problem, the paper uses a method based on serial images analysis. The vehicle speed has been measured accurately by the analysis of two frame images between which there is a certain time interval. The error of speed measurement has been analysed.The buried induction loop has the problem of complicated fixing, easily damaged and so on. Aiming at the problem, the paper uses a measurement method of vehicle flux by simulating the work principle of buried induction loop. A certain dummy detection area on driveway is setted. Vehicle flux has been measured by detecting the change of grayscale.The above algorithm has been realized by experiments. The experimental results show that the algorithm has the advantage of high precision and less calculation. It has a good practicality in toll stations on freeway, automatic charge parks and so on.