Occlusion Adaptive Motion Estimation Techniques for Aerial Object Tracking: A Comprehensive Review

Aerial object tracking plays a crucial role in applications that span from surveillance and reconnaissance to autonomous navigation and environmental monitoring. One of the most critical challenges in aerial tracking systems is occlusion, which can significantly degrade tracking performance and result in complete track loss. This paper provides an extensive review of occlusion adaptive motion estimation methods tailored for aerial object tracking applications. It systematically reviews the transition from conventional correlation-based techniques to cutting-edge deep learning methods, comparing their performance in tackling different occlusion situations. The review covers basic motion estimation concepts, occlusion detection processes, adaptive tracking processes, and performance assessment methodologies. Based on a critical evaluation of available literature, we recognize ongoing research directions, point out ongoing challenges, and suggest future directions. The results suggest that hybrid methods interweaving various combinations of motion estimation algorithms with enlightened occlusion treatment exhibit better performance in demanding aerial scenarios, although computational overhead is still a limiting factor for real-time applications.