Video Detection Algorithm Research And Practice Of Traffic Incident

Traffic incidents detection system plays an important role in the intelligent transportation system. There are many methods to detect the traffic incidents and the video detection method is one of the most rapidly developing technologies in recent years. Due to installation convenience, large detection coverage and high effectiveness, the video detection method is becoming a hotspot in the research field of intelligent transportation system.In this paper, we study several vehicle detecting and tracking methods and propose some novel algorithms aimed at the implementation of traffic incidents detection. Our works mainly focus on the following aspects:[1] Detection of the moving vehicles. We use two different methods to initialize the background first and then update it dynamically. By the background subtraction, dilation, erosion and connectivity judgment method, the vehicles can be located. The experiments show that our algorithms'performances are satisfactory.[2] Tracking of the moving vehicles. We propose a tracking algorithm which based on the time and vehicle template characteristics to track the vehicles, which combines the vehicle location computed during the detection of vehicles. Meanwhile, we provide different strategies to solve different cases such as the disturbed vehicle, the incomplete vehicle shape, and so on. At last, we realize the detection of converse and vehicle flow by using tracking information. The experiment results show that our algorithm works effectively and stably in the vehicles tracking and the detection results are accurate.[3] Detection of crossroads traffic incidents. In view of the real time requirement of system and complex condition of crossroad, we propose a detection algorithm based on the virtual loop and tracking of the rear part of the vehicle. This method can detect and judge traffic incidents such as vehicle flow, vehicle speed, rushing the red light, over speed, etc. The experiment results show that the algorithm satisfies the real time requirement and detects traffic incidents with high accuracy.