Visual object tracking based on improved particle filter

Abstract:

Aiming at the accuracy and robustness requirements of target tracking algorithm, we propose a visual target tracking algorithm based on improved particle filter. First, the target appearance model is described by establishing a variety of features, and the weighting coefficients of each feature component are adaptively adjusted. Then, we exploit the classification resampling method to solve the problem of particle degradation and scarcity in the original resampling method. Finally, a new template updating mechanism is proposed, which can adaptively select moving templates or original templates. The experimental results demonstrate that the improved algorithm has good tracking accuracy and robustness on the challenging tracking video sequences, and it can cope with the complex conditions such as low resolution of video images, rotation change of target, partial occlusion and so on.