Design and Experimental Evaluation of an Operator-in-the-Loop Visual Tracking and Autonomous Navigation System for Fixed-Wing UAVs

Abstract

This article presents the design and implementation of a kamikaze mission framework for a fixed-wing flying-wing unmanned aerial vehicle (UAV) equipped with an operator-in-the-loop ground control system. The proposed approach enables real-time target designation from an onboard Raspberry Pi camera stream operating at 30 FPS, where the operator selects the desired object through a graphical interface. Following selection, a vision-based tracking module utilizing optical flow techniques implemented in OpenCV establishes continuous target lock independent of object class or appearance. The tracked target maintained within a dynamically updated region of interest, providing persistent guidance information throughout the flight. Upon operator authorization, the UAV transitions into a terminal guidance mode and autonomously executes a dive maneuver toward the designated target. The system was first validated in a RealFlight simulation environment and subsequently evaluated through field experiments conducted at multiple altitudes (100 m, 300 m and 500 m). Experimental results demonstrate impact accuracies of 6.4 m, 8 m, and 4.8 m, respectively. The proposed architecture integrates onboard visual processing with ground-control operations, offering a class-agnostic and flexible targeting capability. The results confirm the effectiveness of combining operator-driven target selection with real-time optical tracking for reliable terminal guidance in UAV-based mission scenarios.