Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.538447
Title: Robust airborne 3D visual simultaneous localisation and mapping
Author: Nemra, A.
Awarding Body: Cranfield University
Current Institution: Cranfield University
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
The aim of this thesis is to present robust solutions to technical problems of airborne three-dimensional (3D) Visual Simultaneous Localisation And Mapping (VSLAM). These solutions are developed based on a stereovision system available onboard Unmanned Aerial Vehicles (UAVs). The proposed airborne VSLAM enables unmanned aerial vehicles to construct a reliable map of an unknown environment and localise themselves within this map without any user intervention. Current research challenges related to Airborne VSLAM include the visual processing through invariant feature detectors/descriptors, efficient mapping of large environments and cooperative navigation and mapping of complex environments. Most of these challenges require scalable representations, robust data association algorithms, consistent estimation techniques, and fusion of different sensor modalities. To deal with these challenges, seven Chapters are presented in this thesis as follows: Chapter 1 introduces UAVs, definitions, current challenges and different applications. Next, in Chapter 2 we present the main sensors used by UAVs during navigation. Chapter 3 presents an important task for autonomous navigation which is UAV localisation. In this chapter, some robust and optimal approaches for data fusion are proposed with performance analysis. After that, UAV map building is presented in Chapter 4. This latter is divided into three parts. In the first part, a new imaging alternative technique is proposed to extract and match a suitable number of invariant features. The second part presents an image mosaicing algorithm followed by a super-resolution approach. In the third part, we propose a new feature detector and descriptor that is fast, robust and detect suitable number of features to solve the VSLAM problem. A complete Airborne Visual Simultaneous Localisation and Mapping (VSLAM) solution based on a stereovision system is presented in Chapter (5). Robust data association filters with consistency and observability analysis are presented in this chapter as well. The proposed algorithm is validated with loop closing detection and map management using experimental data. The airborne VSLAM is extended then to the multiple UAVs case in Chapter (6). This chapter presents two architectures of cooperation: a Centralised and a Decentralised. The former provides optimal precision in terms of UAV positions and constructed map while the latter is more suitable for real time and embedded system applications. Finally, conclusions and future works are presented in Chapter (7).
Supervisor: Aouf, N. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.538447  DOI: Not available
Share: