Research On Traffic Information Detection Based On Video Image Process

Traffic information real-time monitoring and analyzing is more and more important to strengthening management of the traffic in the city. In order to collect real-time traffic information, video based detection method is valued more and more important. By placing above the highway, the video camera can obtain a lot of traffic parameters. These parameters have the significant meaning to the traffic automation management. Video based traffic detection technology is becoming the mainstream in the field of traffic information collection.First, this paper research on the traffic parameter detection technique based on regionalinduction, mainly discusses the virtual detection line. This vehicle detection is based onroadway which sets two virtual lines, like the inductive loop sensor. The key idea of thismethod is converting the 2-D digital image information to 1-D time signal by virtual linepreprocessing, which reduce the computational load.There is much interference in the real traffic scene, which reduce the detection accuracy seriously. This paper represents a background initialization and updating method, which increase the veracity of foreground object detection. Pedestrians and night vehicle headlight are the most important interference. By using the object moving direction estimation, the interference of pedestrian and night vehicle headlight is removed. A shadow removal method is used to decrease the shadow interference. Experiment on different traffic scene and analyzing the shortage of virtual line technology, a scheme that can solve the false detection and miss detection which caused by vehicle driving on two roadways and driving too near. The experimental result shows that this system has a good adaptability, high detection precision and good real-time.Otherwise, in order to solve the problems of driving across the line and swerving, this paper research on the object tracking based on particle filter, which is superior on nonlinear and occlusion issues. The model which composed of the object centroid and movement parameter is used to tracking. Experiment on pedestrian with occlusion and vehicles with complex movement mode, the results show that the particle filter is adaptive to occlusion and has a good performance on vehicle tracking. The object tracking based on particle filter can make up for the shortage of virtual detection line technology.