Research On Vehicle Flow Detection Based On Computer Vision

With the development. of technology, computers are widely used in all kinds of fields. Computer vision is to use imaging system to take place of eyespot as input method and process and interpret image in place of human being brains. In this paper, focusing on the application of surveillance in real-time vehicle flow, we talk about the whole process of detecting vehicle flow using computer vision and put emphasis on real-time tracking of vehicle. We first reconstruct, the background of scene from sequential input image, then subtract the background reconstructed from real-time input image. We use Otsu arithmetic to binary the different image, it work well on weak image, and then dilation operator is used to improve the discontinuity of moving target, following that section of line coding method is used to analysis the connectivity of line and segment the image, then we get moving targets, after that, object merge is used to accurately segment target further, later α-β-γ filter is put up to predict and track target. Based on the above method, we can improve the track of moving target accuracy. Experimonfal results show that the algorithm is effective and fast operating speed.This paper is consists of six parts. The first part introduces the development of computer vision and the application in intelligent transport system, compares the different methods about vehicle flow detection and moving target tracking. The second part lays emphases on the technology of reconstruction of background from sequential real-time image. The third part talks about the pretreatment of thedifferent gray image, especially binary the image with weak target and the morphologic filter of image. Line segment coding is put up in the fourth part, we bring forward a fast method to segment targets and get, their feature at the same time, and compare it with normal algorithm and find it is more fast. In the fifth part we describe the procedure of amalgamating target and put emphasis on the research of target tracking method, by building up the moving status of target we can solve the problem of occultation. α-β-γ filter is adopted to predict next position of target so we can reduce the range of seeking matched target. The sixth part describes the framework of the system, also discuses the detection of parameters in occult traffic.