Siamese object tracking for unmanned aerial vehicle: a review and comprehensive analysis

Unmanned aerial vehicle (UAV)-based visual object tracking has enabled a wide range of applications and attracted increasing attention in the field of artificial intelligence (AI) because of its versatility and effectiveness. As an emerging force in the revolutionary trend of deep learning, Siamese networks shine in UAV-based object tracking with their promis -ing balance of accuracy, robustness, and speed. Thanks to the development of embedded processors and the gradual optimization of deep neural networks, Siamese trackers receive extensive research and realize preliminary combinations with UAVs. However, due to the UAV's limited onboard computational resources and the complex real-world circum-stances, aerial tracking with Siamese networks still faces severe obstacles in many aspects. To further explore the deployment of Siamese networks in UAV-based tracking, this work presents a comprehensive review of leading-edge Siamese trackers, along with an exhaus-tive UAV-specific analysis based on the evaluation using a typical UAV onboard processor. Then, the onboard tests are conducted to validate the feasibility and efficacy of representa-tive Siamese trackers in real-world UAV deployment. Furthermore, to better promote the development of the tracking community, this work analyzes the limitations of existing Sia-mese trackers and conducts additional experiments represented by low-illumination evalu-ations. In the end, prospects for the development of Siamese tracking for UAV-based AI systems are deeply discussed. The unified framework of leading-edge Siamese trackers, i.e., code library, and the results of their experimental evaluations are available at https:// github.com/vision4robotics/SiameseTracking4UAV.

Automated summary

UAV-based visual object tracking using Siamese networks is versatile and effective, but it faces obstacles due to limited computational resources and complex real-world circumstances. This study provides a comprehensive review and analysis of leading-edge Siamese trackers, evaluates their feasibility and efficacy through onboard tests, identifies limitations, and discusses the prospects for the development of Siamese tracking in UAV-based AI systems.