Research And Implementation Of Traffic Incident Detection Base On PTZ Camera

Moving object’s detection and tracking is the most important part for intelligent surveillance system. It has made many achievements after years of research. However, due to the complexity of real-world environment factors, so moving object’s detection and tracking is still an important part of the study.In this paper, the main research content is the road incident detection method base on the PTZ (Pan Tilt Zoom) camera. In the road incident detection, the main research content is the vehicle detection and tracking method. In view of the existing literature and the OpenCV for vehicle detection and tracking effect is not very ideal in a complex environment, such as the changeable weather, larger vehicles shadows, rainy day, dim light and night. This paper chose to solve two aspects of content, including the larger vehicles shadows and the night. And after the PTZ camera movement, the problem of the road marking was damaged is solved. This paper mainly has the following several aspects of work.First, shadow detection methods are studied, the purpose is to remove the interference of the shadow. In this paper, some classic shadow detection method is studied and implemented. According to the deficiency of these methods of the shadow detection, this paper design a shadow detection method, which combined with based on HSV color space with based on the gradient, and blending the idea of blob in it.Second, the vehicle detection and tracking methods are studied in the night. In the study of the vehicle detection in the night, this paper improved the method of the vehicle’s light detection, so it can adaptive to detect the vehicle’s lights. In the study of the vehicle tracking in the night, a kind of vehicle tracking method is designed. First, tracking a single vehicle’s light, then match it with other single vehicle’s light that have similar characteristics, tracking it as a new object.Third, in view of the problem that the lane line will be destroyed, when the PTZ camera movement. Therefore, in this paper, a kind of the lane line recovery method is designed. The purpose is to restore the original lane line after the PTZ camera movement. The implementation of the method is based on image mosaic.Finally, according to the vehicle tracking trajectory and the lane line, some road incidents will be judged, such as the vehicle retrograde, parking, speeding, slow, etc. Then this paper’s algorithm is implemented by the OpenCV, and it will be compared with the existing tracking framework of the OpenCV.Experimental results and comparison with the existing tracking framework of OpenCV have shown that the proposed tracking framework is effective.