<p>Unmanned aerial vehicles, or UAVs, have become a cornerstone for multi-target tracking in dynamic and unpredictable environments. Despite advancements in methodologies, the approaches have grave shortcomings even in scenarios involving rapid movement of the target, occlusion, and adversarial strategies. These challenges, however, require this research work to propose a novel framework that integrates the Pursuit-Evasion Game Model with Deep Reinforcement Learning for efficient and adaptive multi-target tracking. This framework uses onboard cameras and LiDAR-based capabilities to ensure proper multi-sensor fusion with accurate environmental awareness capabilities. Its robust real-time data processing also saves room for noise targeting, which is provided by the approach employed by IAEKF in this work. Normalized Gamma Transformation-based CLAHE reduces the contrast in preprocessing and adaptive thresholding using Improved Adaptive Weighted Mean Filter filters out the target. Dynamic path planning and obstacle avoidance of the Energy Valley Optimizer enables it to identify optimal grid-based paths. Pursuit-evasion game: The d policy model implements the pursuit-evasion game between UAVs and evasive targets, allowing UAVs to predict and counter evasive maneuvers from the target. Besides, the segmentation and fusion approaches efficiently put multiple perspectives into a coherent operation framework effective for object tracking. The system assessment in terms of tracking accuracy, response time, success rate, and energy efficiency has been discussed in the previous sections. Results present the adaptability of the approach to complex scenarios with high success rates and robust performance under adverse conditions, thereby providing a scalable and robust solution for multi-target tracking in UAV operations.</p>