Adaptive Object Tracking System Using Swarm Intelligence and Meta Learning Optimization

Abstract

The core function of computer vision research is object tracking but problems with illumination changes and object obscuration and sensor noise and sudden object motion remain obstacles to reach optimal tracking results. A novel dual-object tracking system developed this study merges Kanade-Lucas-Tomasi (KLT) optical flow tracking and Particle Swarm Optimization (PSO)- based swarm intelligence tracking with meta-learning to boost tracking accuracy and adaptability. This system integrates KLT for tracking target dominant points across video frames with PSO to concurrently track boundary information which produces strong object localization results in dynamic conditions. The system benefits from meta-learning because it helps track procedures by establishing universal knowledge across different scenarios while learning new tasks from limited training data. The experimental testing demonstrates how the proposed system achieves superior performance over current tracking methods in terms of both precision and operation speed and stability. The system demonstrates effective performance when tracking short and long videos under static and dynamic background conditions while overcoming deep learning tracking limitations. This study contributes a cost-effective, scalable, and adaptive tracking framework, suitable for applications such as autonomous navigation, surveillance, and human activity recognition, paving the way for future advancements in intelligent object-tracking systems.