Research On Real - Time Vehicle Queue Length Video Detection Technology

With the rapid development of Chinese economy, the number of cars in China also increases remarkably. The traffic problems caused by the rapid surge in the number of vehicles are an important issue that the government faces. Recently, with the progress of Machine Vision Pattern recognition technology, the Intelligent Transportation System has reliable technical support, which gradually becomes an inevitable trend in the development of today’s traffic. The research of vehicle queue length based on the video analysis is a good way to solve practical research on the time of planning the traffic lights and an important research direction of Intelligent Transportation Systems. Also, there is a broad application prospect for the future of urban road traffic system. This paper makes an intensive study of technology of the vehicle queue length based on video analysis of Intelligent Transportation System to do a thorough research. Following is the main work.(1) Video image preprocessing. According to the actual situation, we compared several filters how to deal with two common Gaussian noise and Impulse noise. The experiment results showed that the mean filter and Gaussian filter can be eliminated relatively good Gaussian noise, median filter can be better to remove Impulse noise.(2)Lane divided. When do vehicles queue length measurement, the cars on a driveway easily influence to the other lane, which impact the queue length measurement accuracy. For the above problem, this page firstly detects the vehicles of one intersection lane and solves the problem of the vehicle’s interference of the next lane through dividing each lane by the detection of each lane. For the detection of lane, bilateral thresholding and morphological closing operation filling are used to solve the problem of interference of obstacles on both lanes. Then hough probability and linear fitting are used to the detection of lane. Experiments can show the good detection results.(3)Measuring the length of the vehicle queue. To determine the parking status when the vehicle queue, this paper presents a detection algorithm based on combination of ViBe and inter-frame difference method.To measure the length of the vehicle queue, a detection algorithm which combines the fixed frame and telescopic frame is proposed. Fist, the parking status inside the fixed frame is detected. Then, ensure the front and measure the length of the vehicle queue by the rear detection. The telescopic frame size is variable and it’s location is constantly moved from a fixed frame back along the lane area virtual wireframe. Through the fixed line box detects the vehicle is determined queued state, mobile telescopic frame to determine the queue length of the vehicle. Experimental results show that the proposed algorithm for measuring the length of the queue of vehicles have good results.