Research Of Traffic Parameters Detection Technology For Intelligent Transportation System

With the fast development of economy and the process of urbanization in our country, the quantity of motor vehicle grows rapidly which makes the people's life more convenient and worsens the road traffic at the same time, such as the frequent traffic congestion and also more traffic accidents. In order to release the pressure of the transportation, improve traffic capability and reduce the traffic accident, this paper develops a vehicle flow detection technology for urban district on the base of computer vision. The key to this system is the detection of the lane and the vehicle, which are mainly researched. The main work of this paper are as follows:1. After anglicizing several methods used in background extraction, the background is acquired using method of mean value. Based on that, the moving target is extracted by adopting background difference. And a straight model is constructed to describe the lane mark; Based on the white mark, lane detection is performed by employing the color image segmentation, edge extraction and Hough transform.2. After finishing the lane detection, a horizontal projection method for vehicle detection based on the white lane marks is adopted. According to the horizontal projective histogram of foreground objects, we can judge and confirm the existence of a car by checking the pixel characteristics of horizontal projection.3. This paper tracks the vehicles through image sequences by the repetition of the"matching-correction-prediction"strategy and calculates traffic parameters after finishing the tracking. First of all, corner point of the moving vehicle is detected, and according to the hypotheses of motion coherence, every part of the car is at the same speed. Kalman filtering is employed to predict the position of the corner in next frame of the image sequence, which reduces the search region when calculating new corners and consequently improves the operation speed of the system. The real-time characteristic of vehicle tracking is satisfying.Experimental results show that the lane detection and the moving object detection and tracking algorithm in this paper are robust, efficient and suitable for real-time processing.