An Accurate and Robust Method for Absolute Pose Estimation with UAV Using RANSAC.

Area of Science:

• Robotics and Computer Vision
• Geospatial Navigation and Mapping

Background:

• Accurate absolute pose estimation is crucial for Unmanned Aerial Vehicle (UAV) navigation and mapping.
• Traditional methods using artificial control points are often time-consuming and may lack accuracy with limited data.

Purpose of the Study:

• To develop an accurate and robust method for UAV absolute pose estimation.
• To address challenges of time-consuming data acquisition and low accuracy in existing methods.

Main Methods:

• A customized UAV equipped with a light source (as a 3D point) and Real-Time Kinematic (RTK) for precise positioning.
• Utilizing feature extraction for 2D-3D point correspondences and Random Sample Consensus (RANSAC) to mitigate outliers.
• A novel pose refinement technique employing an error transferring model, gradient descent, and normalization to optimize reprojection error.

Main Results:

• The proposed method demonstrates superior numerical stability, noise sensitivity, and computational speed compared to state-of-the-art solvers.
• Effective outlier removal and pose refinement achieved even with a high outlier ratio.
• Successful validation using both synthetic and real-world image data.

Conclusions:

• The developed method offers a significant advancement in accurate and efficient UAV absolute pose estimation.
• The customized UAV system provides a practical solution for obtaining mass 2D-3D point correspondences.
• The approach is robust and performs well under challenging conditions with significant data noise and outliers.