An Augmented Reality Geo-Registration Method for Ground Target Localization from a Low-Cost UAV Platform

Area of Science:

• Robotics and Automation
• Geospatial Information Science
• Computer Vision

Background:

• Accurate geo-registration of Unmanned Aerial Vehicle (UAV) video is crucial for target localization.
• Low-cost UAVs often have inaccurate attitude data from Inertial Measurement Units (IMUs), limiting geo-registration precision.
• Magnetic compasses in UAVs are susceptible to environmental interference, leading to yaw errors and registration inaccuracies.

Purpose of the Study:

• To develop an augmented reality-based method for precise geo-registration of videos from low-cost multi-rotor UAVs.
• To enhance the accuracy and stability of 3-axis body attitude data by merging Real Time Kinematic Global Positioning System (RTK GPS) and IMU data.
• To propose a target localization method that determines ground target coordinates within the UAV video stream.

Main Methods:

• An extended Kalman Filter (EKF) model was employed to integrate RTK GPS data with onboard IMU data.
• A modified EKF was utilized to combine RTK GPS and IMU data for improved geo-registration and target localization.
• The proposed geo-registration model was used for accurate determination of ground target coordinates from the video stream.

Main Results:

• The integration of RTK GPS data significantly improved the accuracy and stability of the 3-axis body attitude data.
• The modified EKF approach enhanced the precision of both video geo-registration and ground target localization.
• Localization results demonstrated practical accuracy when compared against reference coordinates from satellite imagery.

Conclusions:

• The proposed augmented reality-based method effectively addresses the limitations of low-cost UAV attitude data for geo-registration.
• The fusion of RTK GPS and IMU data via an EKF provides a robust solution for accurate UAV video geo-registration and target localization.
• This research offers a practical and accurate approach for geospatial applications utilizing UAV video streams.