Research Of Traffic Incident Detection Algorithm And Its Implementation On DSP

In recent years, with the rapid development of the national economy, Car usage is growing rapidly, and traffic violations and accidents also increases significantly. The demand of manual monitoring for traffic incidents such as illegal parking, vehicle accidents and vehicle retrograde has been unable to meet. It is against this backdrop, the traffic incident automatic detection system, which is based on machine vision, is brought into being.And such detection system has played a key role in intelligent and automatic traffic video surveillance systems. The development and application of Automatic Incident Detection system can achieve the goals of efficiency, accuracy and high rate, which can provide convenient conditions for such as rapid processing for events, reduction of traffic delays and the avoidance of two times traffic accident.By reading a large number of reference papers from domestic and foreign, and getting a basic acquisition of traffic incident detection algorithm research status, summed up the video-based determination of typical traffic information, this paper designed the moving vehicle segmentation, location and tracking algorithm for real-time image. Based on the results of vehicle tracking, illegal parking, vehicle retrograde, traffic accidents and other traffic incidents can be detected by this algorithm. This major work completes as follows:In the aspect of algorithm research, firstly fulfill the goal of foreground segmentation and target location separately using the Gaussian mixture model-based background subtraction method and the method which combines background subtraction and frame subtraction algorithm. After comparing the results of both two methods, the choice of using the method that combines background subtraction and frame subtraction algorithm is determined and the algorithm will be implemented on the DSP. Secondly, the use of image erosion, dilation and connected domain algorithms can filter out interference region and the smallest enclosing rectangle of the moving target. Thirdly, moving target tracking and vehicle trajectories extracting. Finally, a vehicle tracking trajectory based traffic incident detection method was given, which realized the detection of the traffic events such as illegal parking, retrograde and the parking caused by traffic accidents.In terms of hardware implementation, firstly, this paper analyzes the hardware requirements, and selects ICETEK-DM6437-B evaluation board as a hardware platform and the CCD camera as a video capture tool; Secondly, by automatically extracting the lane to determine the target area, and down-sampling the sampled image to reduce the amount of data processed by the DSP, the efficiency of detection system is significantly improved; Finally, The algorithm is completed in CCS integrated development environment, using the C language.Though testing, the system can effectively achieve the traffic incident detection of the dynamic images, have a certain flexibility and reliability, occupies a less resource, have a fast rate, and have a extensive application prospect.