Research Of Vehicle Identification Technology Based On Video Surveillance

Intelligent Transportation Systems is a hot research area in the field of traffic video surveillance. It represents the trend of the future traffic management. It has been considered as the effective way to solve the problem of road congestion, frequently accidents, to ensure smooth and safety traffic, and to maximize the capacity of existing roads. vehicle detection and speed determination are the key technologies on the bottom of ITS, they are the source data of entire system. The detection accuracy of them have a direct impact of the results of the subsequent target tracking, feature extraction, and behavior understanding, so they play a decisive role in the ITS. Widely used methods based on monocular vision use gray or color information to build the mixture gaussian background model for the moving object detection, and the key step of detection is the building and updating the background model. But methods based on monocular vision are sensitive to light and have problems of black hole and shadow. It is hard to update the background model adapting to light changing, thereby it is easy to cause the problem of false vehicles detection.Based on analysis of others’previous research and exiting problems, the paper which focus on vehicle detection and speed determination, have a deeply study on some key technologies of image preprocessing, updating of background modeling, segmentation of foreground and background, post-processing of moving detection and speed determination. Based on the characteristics of the traffic scene, this paper proposed a binocular vision based mixture gaussian moving object and speed detection methods. Experimental results show that our method obtained accurate contour and speed of the moving object and achieved the state-of-the-art performance.The mainly work of this paper is composed of two parts by vehicle detection and speed determination. Following content is to introduce them:In vehicle detection the paper proposed a new Gaussian mixture background modeling method based on binocular vision. This method utilizes the depth information of binocular vision to establish a background model. It solves the problem based on the monocular visual motion detection such as the black hole, the frequent changes of shadows and light, it reduce the false detection rate of the moving object, and also improve the accuracy of the contour of the moving target.In speed determination based on the results of segmentation and related depth information the paper proposed a novel speed detection algorithm which can adapt to a variety of space shooting angles. It successfully solved the problems based on the monocular vision such as the difficulty to obtain the actual distance of height, angle of camera and actual distance in image flow.