Research On Target Tracking Algorithm Under Complex Environment

The research of moving target tracking in the complex environments is a hot topic in computer vision. It has a wide range of applications in military control and guide, intelligent traffic, medical diagnoses, human-computer interaction and so on. Therefore, it is worthy of learning the moving target tracking algorithm in the complex environments. But, it is difficult to track the moving target as a result of the target change and the complex background interference.In this thesis, the research is focused on the tracking algorithm of single target. And a tracking algorithm is proposed which is based on the improved resampling particle filter. In many target tracking algorithms, particle filter algorithm has good performance in nonlinear non-Gaussian system, and the moving target system is a typical nonlinear non-Gaussian system in the complex environments. This thesis studies the basic principle of PF and introduces the improved resampling particle filter. On this basis, this thesis proposes a new resampling algorithm which is based on the particle weight and the state space information by combining the deterministic of resampling algorithm. We designed a simulation experiment to test the performance of the improvement resampling particle filter. And the result shows that the improved resampling method maintains the diversity of the particles and improves the quality of the particles and the accuracy of the particle filter. Then, this thesis studies how to build the target observation model. For the complex environment, we design a target observation model which is represented by the shape and color cue, and the automatic regulation mechanism about the weights. To improve the robustness of the observation model, the weights of shape cue and color cue are adjusted according to the tracking result of the previous frame.This thesis designs and accomplishes the improved moving target tracking algorithm by combining the new target observation model and the particle filter with the improved resampling process algorithm. And there are some experiments for the improved tracking algorithm. The results of experiments show that this improved moving target tracking algorithm has a better performance in stability and accuracy than the sampling importance resampling algorithm. And the results of the tracking experiments in the complex environment show that this improved moving target tracking algorithm has good performance when the tracking target rotation, illumination, partial occlusion ect. These results prove that the tracking algorithm is suitable for the moving target tracking in the complex environment.