UAV Control for Close Tracking of a Flying Object Using Search Region Focused Detector and Target-Visibility Enhancing Control

• Published in: IEEE access Vol. 11; pp. 139326 - 139334
• Main Authors: Lee, Donghee, Park, Wooryong, Yi, Junhak, Byun, Woohyun, Huh, Soobin, Nam, Woochul
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Autonomous aerial vehicles; chasing flight control; Computational modeling; Detectors; Drone aircraft; Drone vehicles; Drones; Field of view; Flight; Public safety; Radar tracking; Remote sensing; search region focused detection; Target tracking; target-visibility enhancing control; Tracking control; Unmanned aerial vehicle (UAV); Unmanned aerial vehicles; Visibility; Visual flight; Visualization
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2023.3340234

As drones become more prevalent, their malicious use threatens public safety. To reduce this threat, an unmanned aerial vehicle (UAV) that can chase unknown drones is required. However, visual detection for the chasing flight is challenging because the target objects can be small and fly at high speed. Additionally, if the background in the UAV vision is similar to the color of the flying object, detection becomes more difficult. Thus, this paper presents a new UAV system to effectively detect and chase randomly flying objects. First, the detecting performance of YOLOv8 was improved by introducing a search region (SR) technique, which can reduce the probability of detection failure and increase the inference speed. Second, the UAV is controlled to visually capture its target from a different field of view to improve target visibility. If the target object is less noticeable within the image background, the UAV moves to change its position with respect to the target considering the covariance value of the target and the background pixels. The performance of the proposed control system was validated with outdoor flight experiments, where the UAV successfully chased an object that was flying randomly at a speed of 5 m/s with an average distance error of 0.66 m.