Robust online tracking using orientation and color incorporated adaptive models in particle filter

Moving object tracking has received much interest in the field of computer vision due to the increasing need for automated video analysis. Particles Filter is a very promising object tracking method since it is suitable for non-linear and/or non-Gaussian applications. Most particle filter applies color information in target model which might fail in the presence of similar colored objects in the scene. This paper presents the integration of color and orientation features in particle filter to make full use of the distinctive target features. Also, an improved Gaussian weighting function for target models and an updating scheme with an adaptive updating ratio are proposed. The proposed approaches are applied in the real-time video sequences with different occlusion conditions to test the robustness of the proposal. Experiment results show that stable and accurate tracking performance is achieved even when the target is occluded by a similar colored object.