Visual Tracking via a Novel Adaptive Anti-occlusion Mean Shift Embedded Particle Filter

Visual tracking is a significant research field in computer vision. Despite the development of numerous algorithms, the challenge of achieving effective visual tracking in dynamic environments persists. Among various methods, the particle filter (PF) excels in visual tracking due to its adaptability in nonlinear and non-Gaussian environments. In this article, a novel adaptive anti-occlusion mean shift embedded particle filter (AAO-MSPF) is presented as a distinctive approach to address complex tracking scenarios. The integration of the mean shift algorithm can significantly elevate particle prediction accuracy within the framework. The incorporation of the modified particle swarm optimization algorithm optimizes particle distribution and significantly improves the tracking performance. Furthermore, the proposed anti-occlusion module utilizes block-based detection to identify occlusion, enabling adjustments to the motion model. This technique contributes to improved tracking performance, distinguishing our method from others. After a comprehensive comparative analysis, the experimental results indicate that the proposed method AAO-MSPF demonstrates robustness and stability under challenging dynamic conditions and surpasses other trackers in tracking performance.