Research And Application Of The Traffic Flow System Based On Video Analysis

In recent years, along with economic development and social progress, the number of vehicles is increasing and the traffic flow is also continuously improving, that causes some impact on the transportation infrastructure, scheduling and management, when people enjoy the convenient transportation, is also facing various problems at the same time, it is need to create a sophisticated intelligent transportation system to meet the development needs. The traffic flow’s automatic recognition and statistical are an important part of intelligent transportation system to achieve intelligent scheduling and management, vehicle detection and tracking technology is the key of traffic statistics. Traditional detection method can not meet the needs of the time, real-time and efficient target detection and tracking method has been a hot research field of computer version, using computer version technology to detect and track moving target has become an important direction of research and development, and become widely used in Intelligent Transportation.This paper researches digital image processing techniques used in traffic flow detecting, the main work includes moving vehicle detection and tracking.The purpose of moving vehicle detection is removing the real background from the image to extract vehicle characteristic regions, this paper uses Gaussian mixture background model which based on the chrominance components of YUV color space and background subtraction to detect the moving target, researches the inhibition of the chrominance components to the shadow, first using Y and U chrominance components of the YUV color space to model the Gaussian mixture background model and update the background, then get rid of the shadow after using background subtraction method, relative to the commonly RGB color space modeling, reducing the amount of computation in the background building. After extract the foreground moving target area, extract the external rectangular area of the target with morphological filtering method and external rectangle extraction algorithm. Moving vehicle tracking technology uses Kalman filtering motion estimation and multiple features matching. There are two stages of Kalman filtering model:prediction, matching and updating, first predict the location of moving vehicle, then achieve the target by using the moving vehicle detection method in this paper, finally match the predicted result with test result and update the model. In this process, judge the traffic flow through the number of model initialization. During the target matching period, uses multi-feature-based matching method, achieves the exact match on moving vehicles through the aspect ratio, the centroid coordinate and other features of the moving vehicles in the image.This study developed a set of traffic flow identification and statistical system based on vehicle detection and tracking technology, and experimental analysis on the highway, achieved some result, provide a real-time dynamic traffic information collection tools for the intelligent transport.