A versatile real-time vision-led runway localisation system for enhanced autonomy

Area of Science:

• Robotics and Automation
• Computer Vision
• Aerospace Engineering

Background:

• Autonomous landing of Unmanned Aerial Vehicles (UAVs) presents significant challenges.
• Reliable runway detection is crucial for safe autonomous UAV operations.

Purpose of the Study:

• To develop and evaluate a vision-led system for autonomous UAV runway detection and landing.
• To enhance the accuracy and real-time performance of runway sideline localization.

Main Methods:

• An onboard computer vision module utilizing Speeded Up Robust Features (SURF) for feature extraction.
• Fast Structured Forests edge detection combined with Kalman filter smoothing for runway sideline prediction.
• Validation using X-Plane 11 simulator and real flight data from the ULTRA UAV.

Main Results:

• The vision-led system achieved an approximate 84.4% accuracy in localizing runway sidelines.
• The system demonstrated real-time performance, outperforming state-of-the-art edge detection methods.
• Effective validation across diverse weather and lighting conditions, including fog and crosswind.

Conclusions:

• The developed vision-led system provides a robust solution for autonomous UAV runway detection.
• The system's accuracy and real-time capabilities significantly aid autonomous landing.
• The approach is applicable to various UAV platforms beyond the ULTRA UAV.