Method And Key Technique Research About Vision-based Traffic Flow Detection

The development of economy brings a lot of convenience into our daily lives, but on the other side it has done lot of damages to the traffic. Fortunately the ITS solves many traffic problems and becomes more and more important in the traffic management。For the very important part of the ITS, the vision-based approach becomes popular in ITS owing to its easy fixing, low expense and high flexibility.The Vision-based traffic flow detection aims at vehicle detection. In this paper a method that combines both background subtraction and frame difference is chosen on the base of the comparison of several methods. The background image is created through several mixture Gaussian models. Shadow detection is considered a lot in this paper. Considering the complexity and accuracy of the detection, we choose the L weight in the CIE Luv color space to model the background and detect the shadow regions. The examination shows that shadows can be exactly detected and the object vehicles can be kept entirely.Vehicles are detected in the foreground images. Image segmentation and feature extraction should be done to the foreground images. We choose the area and contour as the features. Real-time traffic situation is obtained through the traffic detection.The examination shows that shadows can be exactly detected and the object vehicles can be kept entirely. Detection in the CIE Luv color space improves conventional methods. The vision-based method can obtain real-time traffic parameters easily and quickly, and it will become an important method of ITS.