Research On Moving Object Vision Tracking In Complex Environment

For the past few years, visual tracking of moving objects in complex environment is a very important and active research task in the field of computer vision, and it is the core of intelligent video technology. This topic research has important realistic significance, and the results have broad application. This paper presents a thorough study of visual tracking of moving objects in complex environment by single camera in static condition.According to the features of complex environment,the paper analysis factors of affected moving detection algorithm.A method of RGB image background modeling based on the statistical is presented. Background image is constituted by pixels's mean and variance of RGB color channels, pixels's mean means static, the range of pixels's variance means dynamic of background,so background can be adapt to the environment slight interference.The method of background updates based on three levels(global, image region and background variance), which made background can adaptive updates,is designed.For moving detection, an algorithm of improved real-time detecting moving objects, which has the advantages of both frame-difference method and background-difference method to adapt to complex environments, is presented.For visual tracking, an algorithm of multi-objective tracking of Camshift algorithm based on Kalman prediction is presented, because Camshift algorithm based on moving objects'features of color probability distribution has many faults.The created point of the method is that the new moving detected object is labeled a number and created a Kalman filter,which can improve the anti-interference ability of multi-objective tracking algorithm.For application, a hard platform for moving objects detection and tracking based on image grab card and camera and computer is constructed,and a visual tracking soft with visual interface is designed in VC++6.0.Last,a complete system of visual tracking is accomplish by system debug.