In this paper, we propose a visual-based aircraft pose estimation method, and this method can provide positioning and navigation for the aircraft during the autonomous take-off phase. Using images obtained by a camera installed in the aircraft cockpit, the pose estimation model can estimate the horizontal position deviation and heading deviation between the aircraft and the runway centerline. Then, the flight control system uses these two deviations to control the aircraft for autonomous take-off. The pose estimation model is designed by deep learning algorithm, with a simple and efficient structure. We performed 1000 experiments using a Boeing 737 model in a flight simulation environment, and the results show that the maximum horizontal position error and maximum heading error outputted from the pose estimation model during take-off were 1.89 m and 0.95\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^\circ $$\end{document}, respectively. In all experiments, the aircraft was able to complete a well take-off navigated by the pose estimation model.
A Visual-Based Aircraft Pose Estimation Method During Take-Off
Lect. Notes Electrical Eng.
Chinese Intelligent Systems Conference ; 2023 ; Ningbo, China October 14, 2023 - October 15, 2023
08.10.2023
8 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Aircraft recognition and pose estimation
Tema Archiv | 2000
|Aircraft recognition and pose estimation
SPIE | 2000
|