Use this URL to cite or link to this record in EThOS:
Title: Hough transform techniques for recognition and inspection of industrial objects
Author: Cao, X.
Awarding Body: University College of Swansea
Current Institution: Swansea University
Date of Award: 1991
Availability of Full Text:
Access from EThOS:
It is well known that the lack of robustness, reliability and fast processing of computer vision algorithms is a major barrier affecting the wide use of machine vision systems in industrial automation. To this end, the thesis evaluates many Hough transform related techniques and proposes several efficient algorithms suitable for object identification and inspection in industrial scenes. To detect multiple circular objects in a noisy image accurately, two schemes are proposed. The first scheme combines a labelling technique with the adaptive Hough transform for circle detection to resolve the ambiguities in locating peaks in the parameter space. In the second scheme an alternative transform equation is used to generate the parameter space and, for locating a cluster of three points on a circle, a set of rules is developed. Using this approach, the parameter space generated by the transform is more compact that that generated using conventional methods, and hence fast processing can be achieved. To validate the proposed approaches, several experiments are conducted to detect multiple circular objects in both synthetic and real images. The experiments show that the labelling adaptive Hough transform can produce good results when used to detect circular objects in good-contrast images, and the algorithm using an alternative transform equation performs very well, even in images contaminated by noise, with objects touching and/or overlapping each other. An improved technique is also presented for the detection of arbitrarily-shaped objects by incorporating a model-size-reduction scheme into an optimized, generalized Hough transform. For further speed-up of the algorithms developed in the thesis, their parallel implementation on transputer-based networks is investigated.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available