Traffic Parameter Extraction Based On Target Dertection And Tracking In IR Vedio

Intelligent Transportation System(ITS) is the optimal solution for urban traffic surveillance, traffic scheduling and traffic control. And real-time traffic information collection and processing is a prerequisite for intelligent traffic management. So the establishment of a real-time traffic information collection system has its priority for intelligent traffic management. In this paper, for the demand of traffic parameters for intelligent transportation system, a major study based on target detection and tracking in infrared video for traffic parameter extraction problem at the city intersection is expanded.In order to obtain traffic parameters at the intersection, we must detect and identify vehicles at first, then track vehicles'trajectories, according to trajectories of the vehicles to calculate all the necessary traffic parameters at last. The main research contents are as follows:(1) According to the principles and characteristics of infrared imaging, IR properties of urban roads and vehicles are analyzed. We also select a proper infrared image pre-processing method to improve image's SNR.(2) According to the line feature about the structure of urban traffic road, a method utilizing Hough transform to extract road regions is proposed.(3) A novel method based on improved average frame for background initializing and updating is proposed. This method uses the frame difference or background difference to get the target areas and then fulfills those areas with background pixels to suppress the influence of target pixel in frame averaging. Only accumulate a small number of frames, we can generate high-quality background. Based on this background, background subtraction can effectively extract the target region.(4) For vehicle tracking under multi-target environment, a arithmetic combined with GM(1, 1) prediction model and Mean-shift matching is proposed. Using GM (1, 1) prediction model to replace the traditional Kalman filterer associating with the Mean-shift matching method to replace the traditional probability density association can reduce calculation. Simulation results show that the new method can solve the vehicle tracking problem under the condition of "small sample size, poor information uncertainty problem." Based on stable track for the vehicles, traffic parameter extraction methods are studied thoroughly.The last part of the paper summarizes the work and the prospects for future research!...