A comparison of Pose Estimation algorithms for Machine Vision based Aerial Refueling for UAVs

Abstract

This paper focuses on the analysis of the performance of specific 'detection and labeling' and 'pose estimation' algorithms within a Machine Vision (MV)-based approach for the problem of Autonomous Aerial Refueling (AAR) of UAVs. A robust 'detection and labeling algorithm' for the correct identification and sorting of the optical markers is proposed; a sorted list of marker positions is then provided as input to the 'pose estimation' algorithm. A detailed study of the performance of two specific 'pose estimation' algorithms (GLSDC and LHM) is performed with special emphasis on the required computational effort as well as on the robustness and error propagation characteristics. Extensive simulation studies demonstrate the performance of the LHM and GLSDC algorithms and show the importance of a robust 'detection and labeling' algorithm. The simulation effort is performed using a detailed modeling of the AAR maneuver according to the USAF refueling method

Pose-relay videometrics based ship deformation measurement system and sea trials

“…(1) and (2), the attitude and position of the target relative to the reference base can be calculated as long as the attitude and position of each relay station are obtained [18][19][20][21][22][23] via measurement or calibration. Because each relay station is a rigid body, neither the rotation nor the translation of the stations as a whole brought by the installation deformation will affect the final measurement results.…”

Real-time machine-vision-based position sensing system for UAV aerial refueling

“…(1). Within this study, the Gaussian Least Squares Differential Correction (GLSDC) method was used for pose estimation purposes [2,22,23]. The GLSDC is based on the application of the Gauss-Newton method for the minimization of a non-linear cost function formulated in terms of the difference between estimated and detected corners positions.…”

“…(14)(15)(16) was designed to work with a fixed number of m corners. Specific modifications have been introduced for dealing with a time-varying number of corners [23]. Particularly, the nominal corners that are not visible are removed from the estimation process at each time step.…”

Optimal non-iterative pose estimation via convex relaxation

“…This explains why they are often prime candidates in applications involving precise localization and orientation determination, particularly in robotics, augmented reality and defense [16].…”