Research And Application Of Video Detection Algorithm On Running Red-Light Event

Due to convenience installation,large detection coverage,multiple detection parameter and easier evidence preservation, The Red Light Runners detection system based on video develops rapidly in recent years, and becomes a hotspot in the research field of intelligent transportation system. At the same time, because of the complex background of detection: large vehicle flow,mutual interference between moving objects,constantly changing of the light and so on, it makes the correctness of video detection system lower.Owing to the background above, this thesis aims to Red-Light-Runners video detecting algorithm researching and discussing. And propose a novel method Using virtual loop detection technology and vehicle tracking detection technology. Our works mainly focus on the following aspects:[1] According to the characteristics of stop line, this thesis proposes a virtual loop detection algorithm based on the characteristics of stop line. The algorithm implements the detection on moving vehicle according to the judgment on covering stop line. This thesis proposes an algorithm of updating background correspond to the virtual loop. The virtual loop detection algorithm detects efficaciously and quickly and it that can eliminate the influence of shadow.[2] According to the characteristics of having lots of edge information about vehicle tail, this thesis proposes an algorithm of tracking vehicle tail. Linking with virtual loop detection, it implements the detection on vehicles' transgressing-lane behavior. This algorithm can eliminate disturbance aroused by dense vehicles or man, ensure detecting correctly.[3] According to the practice of crossroad, this thesis makes vehicle speed as one of the judge condition that vehicles run red. And proposes an algorithm of velocity measurement based on vehicle tail tracking.[4] This thesis proposes a tracking algorithm based on moving objects' types to deal with the problem of tracking the transgressing-lane vehicles. According to the characteristics of moving object on the detection region, the algorithm distinguishes the different types of the moving objects, and tracking the moving vehicle pertinently. It can improve the speed of the detection and the correctness of detection. According to the parameter of vehicle's barycenter, it also can deal with the problem of hided vehicles, which exists in the detection on vehicles' transgressing-lane behavior.[5] Using virtual loop detection,tracking vehicle tail and tracking transgressing-lane vehicles, the system implements the correct judgment of running red light. This algorithm eliminates the disturbance aroused by heavy vehicle,dense vehicles or man, and improves the system veracity of detection.