| China is a country with frequent occurrence of accidents. In order to reduce andprevent the occurrence of traffic accidents on road traffic, the real-time highway detectionhas been paid more and more attention. So the study of intelligent transportation systemhas also become a hot spot. By the real-time analysis of vehicles and pedestrian on the roadand, it can control the traffic flow and monitor the occurrence of traffic accident.Traffic incident detection system based on the real-time video is a key part of themodern intelligent transportation system. Throughout the real-time traffic image sequencesanalysis and judgment, traffic incident detection system manage to detect the trafficaccident and ensure the traffic environment safely and orderly in time.In this paper, based on actual highway traffic video as a research object，the academicachievements at home and abroad is referred. Throughout comparing with the advantagesand disadvantages of three kinds of common methods of moving target detection, such asframe difference method, optical flow method and background subtraction method andinvestigating the using the conditions of the three detection methods, backgroundsubtraction method is used to detect the moving targets. Also，the theory and the way ofbackground established of the median method, mean method, Gaussian mixture modelalgorithm is introduced. Then，background image of congestion event and parking event isestablished by mean method, and Gaussian mixture model algorithm used for establishingthe background of pedestrian event. Based on the analysis the information of thepedestrians and vehicles on the highway, it implements a video visual traffic detectionsystem to achieve congestion event, parking event, and pedestrian event real-timedetection.This paper uses the Visual C++6.0development environment to simulate thealgorithm of the traffic incident detection. Through experimental tests on actual highwayvideo, it proves that the operator effective and obtain a better result. |
PDF Full Text Request