Motion Estimation by Hybrid Optical Flow Technology for UAV Landing in an Unvisited Area

Area of Science:

• Robotics
• Computer Vision
• Aerospace Engineering

Background:

• Unmanned aerial vehicles (UAVs) face challenges landing in unmapped environments.
• Current landing methods often rely on external infrastructure or markers, limiting autonomous operation.
• Accurate motion estimation is crucial for safe and precise UAV landings.

Purpose of the Study:

• To develop a vision-based motion estimation framework for autonomous UAV landings in unvisited areas.
• To enable UAVs to land without prior environmental information or guiding markers.
• To improve landing accuracy and reliability compared to existing navigation systems.

Main Methods:

• A hybrid framework utilizing optical flow techniques with a multi-scale strategy was implemented.
• The system compensates for the decreasing field-of-view during UAV descent.
• Vision-based motion estimation was employed as an aid for landing performance.

Main Results:

• The vision-aided scheme demonstrated improved landing accuracy compared to commercial Global Positioning System (GPS).
• The framework successfully landed UAVs in new areas without prior guiding marks.
• Sensing errors were effectively minimized by the vision-aided approach.

Conclusions:

• The developed vision-based system offers a robust solution for autonomous UAV landings in unknown environments.
• This approach enhances landing precision and reduces reliance on external navigation aids.
• Future integration with image learning can further expand UAV application versatility.