Abstract:
Target tracking has been already a research direction in the field of pattern recognition and computer vision. Especially, video target tracking has become a focus of study. Target tracking plays an important role in the intelligent monitoring, traffic monitoring, traffic statistics, and so on, due to the variability of shooting environment, the fluid of target motion state, the shading of targets and the interference of likeness, etc. These factors make video target tracking more complicated. Video target tracking can be defined as select single or multiple target from the video sequences filming with cameras, and can give target location accurately and timely, and then get target motion trajectory and its motor habit. Mainly in the framework of particle filter algorithm, this paper analyzes the factors which affect the effect of particle filter tracking, and focuses on the study of the target feature extraction, the similarity measurement method between models, controlling the number of particles, in order to improve the robustness, accuracy and instantaneity of the algorithm. Particle filter video target tracking algorithm based on Mean Shift is proposed. The algorithm uses an iterative process of Mean Shift algorithm after the particle initialization and resampling to reduce the amount of calculation. The algorithm updates the target model after estimating location of the target, in order to adapt to changes of the target. The experimental results show that the effect of the particle filter tracking based on Mean Shift is better, and it needs shorter time.