The Study Of Moving Object Detection And Tracking Algorithm Based On Image Information

Moving object detection and tracking in image sequence is one of the critical issues in the field of computer vision. It merges many technologies such as the science of computer, communication technology, physics, control science and so on. In recent years, it has been widely applied in many aspects including military guidance, safety monitoring, intelligence transportation, medical diagnosis, etc.This paper could be divided into three parts: image preprocessing, moving objects detection and objects tracking, server classical algorithms in those aspects are summarized and profoundly analyzed, such as GMM, Mean Shift algorithm, Kalman Filter, Particle Filter and so on. Based on the studying and comparing those algorithms, a novel algorithm for moving objects detection and tracking is proposed in this paper. GMM is employed for the object detection and improved Marginalized Particle Filter (MPF) is used for objects tracking. MPF merges Particle Filter and classical Kalman Filter, is a powerful tool to solve nonlinear and non-Gaussian problem. The system states of dynamic model are divided into two parts: linear and nonlinear. Kalman Filter is employed for linear state estimation and Particle Filter is used for nonlinear state tracking. On the basis of MPF, this paper employ Mean-Shift algorithm to improve particles sampling. The result of experiment could verify the feasibility and effectiveness of this algorithm. It could meet the real-time requirement.