An unmanned aircraft system (UAS) is an aircraft without a human pilot onboard, controlled either by a ground-based controller or with a pre-set flight plan. With the rise of artificial intelligence (AI), people have become increasingly interested to enable the UAS to navigate on its own, to make autonomous decisions for its flight path while carrying out certain tasks rather than simply executing a pre-set flight path. This requires the UAS to have a certain awareness of its location in space and the obstacles around it. Hence, additional sensors must be mounted on the UAS to give it vision and awareness of its surroundings. However, the payload is still preferably kept to the minimum for better overall efficiency of flight. Here we explore the possibility of reducing payload by combining sensing and navigating tasks together. Specifically, we design the vision-based navigation system to be able to do 3D nominal scene reconstruction through video input from a single onboard 2D camera for object localization and navigation. A nominal scene reconstruction will be sufficient for UAS navigation as it only has to be aware of its proximity to corners and surfaces, to be able to avoid them and not collide into obstructions. Initial works show promising results, i.e. the system can produce 3D point clouds (nominal scene & geometry) from video stream provided by on-board camera.
3D Nominal Scene Reconstruction for Object Localization and UAS Navigation
Lect. Notes in Networks, Syst.
International Conference on Intelligent Autonomous Systems ; 2021 ; Singapore, Singapore June 22, 2021 - June 25, 2021
2022-04-08
14 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
British Library Conference Proceedings | 1997
|Instant Scene Modeler for Crime Scene Reconstruction
IEEE | 2005
|