Autonomous Precision Landing for UAVs Using Sequential Object Detection

This paper proposes an autonomous precision landing system for quadrotor UAVs(Unmanned Aerial Vehicles) using sequential object detection. The system utilizes a single one-axis tilt gimbal camera to perform forward view monitoring during flight and precision control during landing, effectively reducing the computational load of the UAV system. The YOLO(You Only Look Once) algorithm is employed for object detection, where the landing pad is detected at higher altitudes, and an auxiliary marker is additionally detected at lower altitudes to enable more precise positional control. Experimental results demonstrate that the proposed sequential detection approach significantly improves landing accuracy compared to conventional methods that rely solely on landing pad detection. Furthermore, the system maintains consistent precision across various altitudes, effectively mitigating bounding box variation issues at close distances. Verification through simulation and real-world experiments confirms the reliability, accuracy, and practicality of the proposed system, even in resource-constrained UAV environments.