A Height Estimation Approach for Terrain Following Flights from Monocular Vision

Area of Science:

• Robotics and Automation
• Computer Vision
• Aerospace Engineering

Background:

• The proliferation of Unmanned Aerial Vehicles (UAVs) in mapping necessitates advanced perception capabilities.
• Terrain following remains a significant challenge for commercially available UAV systems.
• Existing UAVs commonly integrate cameras, presenting an opportunity for vision-based sensing.

Purpose of the Study:

• To develop a monocular vision-based algorithm for accurate height estimation for UAV terrain following.
• To leverage existing onboard camera hardware for crucial flight altitude data.
• To enhance the reliability and safety of autonomous UAV operations in complex terrains.

Main Methods:

• Utilized optical flow to track visual features from UAV-mounted camera videos.
• Integrated UAV motion data with optical flow for precise flying height estimation.
• Developed a decision tree classifier using optical flow inputs to validate height estimation reliability.

Main Results:

• The height estimation algorithm demonstrated good accuracy in determining flight altitude.
• The decision tree classifier achieved 80% accuracy for positive (trustworthy) classifications.
• The classifier achieved 90% accuracy for negative (untrustworthy) classifications, ensuring robust performance.

Conclusions:

• Monocular vision-based height estimation is a viable solution for UAV terrain following.
• The proposed method effectively utilizes optical flow and motion information for reliable altitude sensing.
• The integrated classifier enhances the trustworthiness of the height estimation system for practical applications.