Study On Algorithm Of Freeway Traffic Incident Automatic Detection Based On Image Analysis

Digital image processing technology is a new science research field which has more development potential, and it is applied more and more in traffic industry. The system of freeway traffic accident detection based on image analysis makes use of the technology of computer vision, artificial neural network, fusion logic, and genetic algorithm, etc. And it has the better efficiency than the traditional traffic flow-based accident detection system. In This thesis, an algorithm of freeway incident automatic detection based on video is proposed. Vehicle detection, morphologic filter, connected region analysis, vehicle tacking and incident detection are included in the algorithm, and the solution for the problem and difficulties in the implementation are proposed. In the shadow detection, an algorithm based the invariability of color space C1C2C3 is proposed. Because the vehicles'running on the freeway can be looked as at a uniform speed in a straight line, a moving vehicle status prediction model based Kalman filter is building, and according to the result of Kalman filter tracking, overlapped vehicles and splited vehicles are determined, template matching based pyramid searching is implemented to overlapped vehicle segmentation; Through straight line fitting, complicated vehicle trajectory is simplified to the collection of lines, and analyzing the change of driving based the angle difference between lines, and the incident is determine. Furthermore, the algorithm based Hide Markov Model is applied to detect vehicles collision. At the same time, the incident detection algorithm is implemented based on VC++ and OPENCV, and it worked on the video recorded on freeway well and proved to be available.