Progress and Perspectives on UAV Visual Object Tracking

With the increasing deployment of unmanned aerial vehicles (UAVs) in security surveillance, disaster response, intelligent patrolling, and wilderness search and rescue, visual object tracking has emerged as a critical perceptual capability. Compared to traditional ground-based platforms, UAVs operate under unique challenges during flight, including drastic viewpoint changes, dynamic background interference, frequent occlusions, and limited onboard resources. This paper focuses on the domain of UAV visual object tracking, providing a systematic review of recent research progress and technological trends. First, mainstream tracking paradigms are categorized by modality, with their evolution in terms of performance, efficiency, and adaptability outlined. Then, based on real-world deployment environments, UAV tracking tasks are analyzed across three representative real-world scenarios—urban roads, aerial targets, and mountainous jungles—with an in-depth analysis of the key challenges and adaptation mechanisms within each setting. Building on this, the performance and capability boundaries of advanced methods are comprehensively compared across five representative datasets under multimodal, multitask, and complex environmental conditions. Finally, the paper discusses critical research challenges and future directions, spanning data foundations, environment-driven modeling, cognitive perception, and other related aspects. This review aims to offer a multi-perspective analytical framework to support the development of next-generation UAV tracking systems with enhanced generalization, robustness, and cognitive intelligence.