Moving Target Detection And Tracking Algorithm Research

Moving target detection and tracking is the main content of machine vision. In the automatic control system, the robot navigation, intelligence and surveillance systems, medical image analysis, video coding moving target detection and tracking have a wide range of applications.The estimated target recursive filtering algorithm is an important cornerstone for tracking system. The famous Kalman filtering algorithm is the best bayesian recursive estimator for solving linear Gaussian for more than 40 years. This paper focuses on the application of partical filtering based on bayesian framework for recursive estimation in target tracking. First, this paper research on the moving target detection algorithm and improve modeling background frame difference method. The method use dynamically updated the template method and divide dynamic image into the foreground zones and the background area.When frame difference have superposition phenomenon, extracting superimposed area and updating the superimposed area into the background model make more accurate background model. Second,in the paper the particle filtering method is applied to target tracking and using dynamic modification search radius methods improve tracking accuracy. In the process of multi-target tracking, label the target block. In the objective existence of occlusion and overlap, the method of calculating the distance between target at first and then comparing between a sub-block displacement and optimal target displacement reduce computational complexity.The last,in the paper for multiple target tracking extracting initial model of the image frame by K clustering method at first and then reading several frames ,using moving target detection methods identify target.