Vehicle Flow Detection Of Intelligent Transportation System

ITS (Intelligent Transportation System) includes mainly computer vision and image processing technology. It can detect and track vehicles from video, without or with a little human intervention. It can offer a more advanced and feasible scheme to deal with traffic accidents. In this paper, focusing on the applications of surveillance in real-time vehicle flow of ITS, the whole process of detecting vehicle flow using computer vision puts emphases on real- time vehicle tracking.This paper design and realize a vehicle flow detection system. The system is composed of image capturing, moving targets detection, morphological filter, joint area detection and moving targets tracking. Some arithmetics are put forward to solve the difficulty in the realization. Firstly, reconstruct the background of scene from sequential input images, emphasizing on intelligently background updating. The brightness of background is different from that of the current image for the change of the weather. Background' s not updating will bring difficulty to vehicle detecting. A new arithmetic is basedon row by row processing and compensation coefficient background real-time updating. Secondly, having compared several threshold segmentation methods, Otsu arithmetic is selected to binarize different images. After that, shadow emerges a part of the target, which affects targets tracking. A new method, which is based on Sobel edge detection operator, is applied to remove target's shadow. Thirdly, after comparing several methods of line connectivity analysis and image segmentation, run length coding method is chosen. Both geometric and color feature are necessary to match moving targets in tracking. A method which is used to extract feature of moving targets while segmenting moving targets by using run length coding method. Fourthly, a - 3 - y filter is put up to predict position of moving targets' center. Color information is used to match targets when several targets merge into one target. Based on the above method, we can improve the track of moving target tracking. Experimental results show that the system is effective and operating speed is fast.