Research On Moving Target Detection And Tracking In Video Surveillance System

With the rapid development of computer vision and communication technology, monitoring technology has undergone tremendous changes. In such a development environment, Intelligent Transportation Systems (ITS) is proposed and given great concern by the world. Intelligent Transportation Systems (ITS), which is combined with artificial intelligence, automation technology, computer technology, information communications technology and some other cutting-edge technology, is aim to form a new powerful real-time, accurate and efficient integrated transport management system. As moving target detection and tracking is the core of intelligent transportation systems, this paper has done some research on this problem. The main contribution and work are described as follows:First, study the development background and trends of the moving target detection and tracking field and then the current domestic and international typical algorithm has been studied and analyzed. Design a set of processes about vehicle detection and tracking in static background and the relevant algorithms has been deeply researched. Then realize the relevant algorithms using VC++software open platform and OpenCV software development kit in the Windows environment.In the pre-processing module, use the video format converter and smoothing filter processing which could reduce the random noise, environmental noise. Focusing on the adaptive Otsu threshold method which is based on color histogram, so that get a better binary image which contains less noise of the moving objects from segmentation image.In the target detection module, first study commonly used detection algorithms such as background subtraction, frame difference, optical flow, and compare their advantages and disadvantages. Then do deeply research on Gaussian mixture background modeling algorithms and built an adaptive Gaussian mixture background modeling combined statistical average background modeling. The target detection module could extract the moving targets successfully.In the target tracking module, in-depth study a tracking algorithm based on region features and kalman filter. First, research an effective region segmentation algorithm for the binary images of the moving target, and then set area, length and width thresholds to further filter out false targets. Second extract the minimum enclosing rectangle, centroid, area and other features and then estimate these use kalman filter, so that the features and search box could be optimum value. Finally, give a matching algorithm using the motion characteristics in consecutive frames.The overall tracking system could continuously track vehicles under these conditions:straight lanes, turning lanes, the size deformation of vehicle, the color of vehicle similar to the environment and the experiment achieves good results.