Use this URL to cite or link to this record in EThOS:
Title: An investigation of a remote visual navigation system for a building inspection robot
Author: Paterson, Alastair Mark
ISNI:       0000 0001 3476 5800
Awarding Body: City University
Current Institution: City, University of London
Date of Award: 1996
Availability of Full Text:
Access from EThOS:
Access from Institution:
The work presented here shows the development of a machine vision algorithm for finding the position of a building inspection robot on the outside of a large building. The reasons for external building inspection are introduced along with the types of tests used. Existing methods are examined giving their limitations in terms of practicality and safety and an alternative using remote access is proposed. The work concentrates on the navigational aspects and shows how one possible solution using machine vision could be implemented and this is compared to similar work carried out elsewhere. The major part of the thesis covers the development of the robot location algorithm starting with the fundamentals of image processing and finishing with the actual robot's position. Different methods of edge detection are investigated and a pixel linking routine is used to group together data in an image that form features and principal lines. The algorithm investigates the use of the lines for detecting vanishing points and tries to identify the features highlighted in the image. The most significant part of the work concentrates on the development of a method of identifying specific features such as a target on the robot and different windows along with a way of matching the features to a computer model of the building thus enabling the position of the robot to be calculated. Results are given showing how the algorithm performed on a model building and robot in the laboratory with various tests using different camera positions, image enhancement and spurious features. The results presented show that the algorithm was capable of finding the position ofa model robot to sufficient accuracy (typically 3% of the size of the robot target) and that the errors measured were predictable. Additional results show how the algorithm performed on a real building and indicate the problems associated with real images with the conclusion that the algorithm will work under a certain range of conditions providing that certain elements of it can be improved.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: QA75 Electronic computers. Computer science