Research On Red Light Violation Detection Based On Video Image Processing

With the increase of urban population and the increase in motor vehicles, the contradiction between supply and demand of traffic has spread from first-tier cities to the second and third tier cities, so improving the strength and efficiency of traffic regulation is imminent. The traditional transportation technology and means no longer meet the requirements of social development, and intelligent transportation systems is considered to be one of the most effective ways to solve the deteriorating urban traffic problems. As an indispensable part of the intelligent transportation system, red light vehicle detection system has important research significance and practical application value. According to statistics, the crossroads are areas of high rate of traffic accidents, most of the accidents are caused because vehicles ran the red light. It is a huge threat to people’s lives and property that vehicles ran the red lights. Therefore, it is very important and valuable to study the identification system to improve road safety through distinguishing vehicles running the red light.This article mainly studies the research on red light violation detection based on video image processing. The work includes the following three aspects:Firstly, in terms of detection of moving vehicles, we use a modified three- frame difference algorithm to detect moving objects. Three- frame difference algorithm has better noise immunity compared to the traditional two-frame difference algorithm, and it can make the outline of the image clearer and get a better extraction effect.Secondly, the paper adopted the feature-based tracking algorithm suited for this system in terms of the tracking the moving vehicle. By extracting the center of mass, the length and width of the external rectangle of the vehicle, using multi- feature fusion method to update a vehicle’s feature information. Then establishing ROI and track list of multiple dynamic vehicle to extract the trajectory of target vehicle.Finally, in the red light behavior recognition, based on the analysis of the trajectory of the vehicle, this paper has put forward specific detection methods in view of the behavior of vehicles running red lights, and given the experimental results on the basis of video material implementation through a red light detection experiments.This paper presents a detection algorithm of vehicle through the red light based on vehicle trajectory analysis. The experiments show that the algorithm could meet the needs of intelligent transportation systems because of strongly accuracy, highly real-time and so on, thus, it has practical application.