Research On Algorithm For Freeway Traffic Incident Effect Prediction Based On Video Detection

With the urban development and the sharply increase of intercity traffic demand, congestion, obstruction during peak increase seriously and traffic incidents occur frequently on the intercity freeway. To solve the problem that the highway incident is apt to cause massive congestion, an effective event detection technology can reduce the occurrence of secondary events and other adverse reactions. Furthermore, a real-time predictive analysis to forecast the impact area of incident can be used as the basis for the incident solution. This paper mainly design and research the freeway incident detection and impact prediction algorithm.Based on the extensive research of the freeway incident detection and impact prediction method from all over the world, a video-based rapid detection algorithm and an impact area of incident prediction algorithm for fast feedback event information were proposed. Firstly, using the smoothing sequence method to obtain the background and using the background differencing method to obtain the foreground target. Then after the shelter detection based on convex hull share, using the modified multi-feature matching tracking algorithm based on kalman filter to track the target vehicle. Secondly, through the detection of the traffic flow speed and flow, using the fuzzy inference method to detect the traffic incident. Finally, built on the traffic incident flow model, calculating the speed of traffic incident flow based on distributed video information and building the vehicle queue dynamic estimation model for future multiple moments to forecast the impact area of incident.In summary, the paper is based on the incident detection method, queue detection method and the traffic flow theory to detect the happen of traffic incident with the video information shooted by the video cameras using knowledge of computer vision and image processing. Then calculate the speed of traffic flow based on distributed video information to forecast the range of incidengt and finally realize the design and achievement of the incident detection and range prediction system. Experimental results show that the method used could obtain accurate prospect information, and it was an applicable and high time-efficiency algorithm for traffic incident detection on the freeway. And the impact prediction method is more accurate and effective compared to some other methods so as to offer basement for management and decision for highway managers.Experimental results show that the method used could obtain accurate prospect information, and it was an applicable and high time-efficiency algorithm for traffic incident detection on the freeway.