Detection and Tracking of Distant Small Moving Targets Based on Particle Filtering and Trajectory Recovery

Moving small target detection and tracking is a classical problem in computer vision, with wide applications in fields such as drone surveillance and traffic monitoring. Traditional methods perform well for medium and large targets but face limitations in small target detection, particularly in low-resolution environments and where features are less prominent. To address these limitations, specialized techniques for small target detection and tracking have emerged, focusing on enhancing feature representation and improving tracking algorithms. However, current models still face challenges with limited generalization in dynamic and complex scenarios, including issues with detection accuracy and tracking drift. This paper proposes an improved particle filtering method that integrates motion information, background modeling, and multi-feature fusion to enhance the performance of moving small target detection and tracking. By leveraging historical motion trajectories for accurate position prediction, and incorporating features such as color histograms and Local Binary Patterns (LBP) to enhance local feature representation, the proposed method optimizes the detection framework, expands the receptive field, and improves tracking stability. To address occlusion, the method utilizes trajectory recovery to ensure continuous tracking. Experimental results demonstrate that the proposed approach exhibits strong robustness and adaptability in complex scenarios, significantly improving detection accuracy and tracking stability.