Autonomous quadcopter navigation in GPS-denied environments, such as aircraft hangar inspection, requires a computer vision system that reliably detects guide paths in real time despite changing lighting. This study evaluates a computer vision system combining HSV color segmentation with Hough Transform as a backup method to detect straight and curved white-line paths, as a preliminary outdoor validation study. The system uses a Pixhawk 2.4.8, Raspberry Pi 5, and Logitech C270 camera, with a dual-ROI algorithm and a PD controller, validated through Gazebo SITL simulation and flight tests at three altitudes (1 m, 1.5 m, and 2 m). Each altitude was tested at least 10 times, with one representative test used for analysis. The best session at each altitude showed the quadcopter completing autonomous takeoff, hover, line following, and landing, with detection rates of 93.24% (1 m), 87.04% (1.5 m), and 90.29% (2 m), and a maximum yaw rate of 24.13°/s during turns. The proportion of detection methods shifted significantly with altitude: HSV accounted for only 1.4% at 1 m but rose to 79.6% at 2 m, with Hough dominating at lower altitudes. The combination proved complementary, improving path detection reliability under the tested conditions.
Copyrights © 2026