| Multiple target tracking is one of the hot issues in computer vision, it is an active research topic which concern the cross-disciplinary including image processing, pattern recognition, probability theory and statistical inference and system state estimation. Multiple target tracking has been widely used in the areas of video surveillance, video compression, and imaging guidance。So far, many target tracking algorithms have been proposed in the field of computer vision. Particle Filter is one of these algorithms, it is based on Bayesian estimation theory and widely used in the area of target tracking. This dissertation studies on the multiple target tracking algorithm based on particle filter. Firstly it sums up the current target tracking techniques and gives brief introduction to Bayesian filtering theory and Particle Filter. Based on the previous work, the major work and innovation of this paper are like the following:1. Under the framework of Particle Filtering theory, we discuss the single target tracking algorithm based on Particle Filter. From the aspects of target movement model, target observation model and resampling algorithm, we introduce the specific methods for realizing the single target tracking algorithm based on Particle Filter, then the tracking results are analyzed.2. A novel graphical model based multiple target tracking algorithm is proposed. This algorithm uses an undirected graphical model for the description of multiple target model, and takes the multiple target tracking problem as the inference of graphical model. Because the Particle Filter can solve the problem of parameters estimation of non-linear and non-Gaussian system, we use it for the inference of graphical model. And we use the Mean Shift to solve the complexity of Particle Filter.3. For the application of face tracking, we propose a face tracking algorithm which based on the integration of two visual characteristics: color and gradient of contour, and it shows good experimental results. |
PDF Full Text Request