3D Positioning of Drones through Images

Area of Science:

• Computer Vision
• Robotics
• Geospatial Analysis

Background:

• Drones traditionally rely on satellite navigation (GPS) for positioning and altitude.
• GPS-denied environments disrupt satellite communication, challenging traditional drone navigation methods.
• Accurate drone positioning and altitude estimation are critical for safe and effective operation.

Purpose of the Study:

• To develop a drone positioning and height determination method independent of satellite signals.
• To evaluate the performance of a proposed multi-scale input network for drone navigation in GPS-denied scenarios.
• To enable single-image-based height judgment for Unmanned Aerial Vehicles (UAVs).

Main Methods:

• Creation of a drone dataset captured at altitudes between 80-200 meters.
• Development and application of a multi-scale input network for positioning.
• Implementation of UAV image-to-satellite image matching for positioning.
• Calculation of UAV flight height using image-derived real-space area and optical camera parameters.

Main Results:

• The proposed positioning index (RDS) achieved 76.3 points.
• Positioning accuracy within 20 meters was achieved at an 81.7% rate.
• The method calculated UAV flight height within the 80-200 m range with a relative error rate of 18.1%.
• Height judgment was successfully performed using single UAV images.

Conclusions:

• The proposed method offers a viable solution for drone navigation in GPS-denied environments.
• Image-based height determination and positioning are feasible without external sensor data.
• This approach enhances drone autonomy and operational capabilities in challenging conditions.