A Number Of Ways Of Moving Target Tracking Algorithm Research

Target tracking technology has been widely used in military, security, traffic monitoring, and other fields. How to track the target interested effectively in the video sequence is a challenging topic in the computer vision. In recent years, many of the classic methods have been used to solve a lot of thorny tracking problems through the efforts of researchers, but at present algorithms only can be effective under some certain conditions or object environment, so it's not perfect. And because of the tracking object itself characteristic's multiplicity with the locating of the environment complexity, enables the target tracking technology to be beset with difficulties in the practical application.First of all, one of the classic methods: Correlation-based method, which is also called template-matching method, is discussed in the thesis. Particularly, we focus on SSDA and MCD, and on the basis of it, the rapid SSDA algorithm which has the function of updating adaptive template is proposed to solve the relevant aspects of tracking problems in real-time, and the influence of the target deformation problem.This is followed by the analysis of the classic Mean shift algorithm. On the basis of this tradition algorithm, several improved methods have been explored. Kalman prediction algorithm is also introduced into the tracking algorithm to solve the issue and block with the interference of similar problems .And HSV color model is introduced into the tracking algorithm as a characteristic to solve the effect which made light changes in tracking.In the last part of this paper, a discussion on the Bayesian estimation of Particle Filter algorithm is made as well as the application of the target tracking, and its characteristics through experiments are summed up. An integrated algorithm is also proposed based on Mean shift algorithm and Particle Filter algorithm. The algorithm can switch Mean shift algorithm and Particle Filter algorithm by measurement parameter adaptively, so that we can enhance the stability of the Mean shift algorithm and enhance the real-time nature compared with traditional Particle Filter algorithm.