Use this URL to cite or link to this record in EThOS:
Title: Object representation and recognition in machine vision
Author: Shneier, Michael Oliver
Awarding Body: University of Edinburgh
Current Institution: University of Edinburgh
Date of Award: 1979
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This thesis is concerned with the representation and recognition of objects by computer. A way of representing objects in terms of primitives, or basic descriptive elements, and relations between primitives is presented and discussed. The representation involves abstracting properties of individual scene elements that serve to describe a class of such scene elements. Similarly, relations that hold for particular instances of objects are generalized to ensure that they will be valid for all instances. By sharing all common primitives and relations, the set of models can be made very compact, and yet preserve many desirable properties. A computer implementation of a version of the representation was developed. It is used to illustrate the descriptive power of the representation. A recognition algorithm is presented that efficiently uses the representation to recognize real world scenes. The recognition involves two passes over the data. In the first pass each scene element is tentatively interpreted as belonging to some subset of the known models, on the basis of matching scene elements with model primitives and performing relational tests. In the second stage the hypotheses are examined, by means of a constraint analysis algorithm that attempts to find the globally best interpretation for each scene element. The value of an interpretation is based on the number of expected relations that were actually found to hold between scene elements for each possible interpretation. Examples of applying the modelling and recognition system are shown in two domains. The primary domain is that of three-dimensional vision, with data provided by a triangulation ranging device. A secondary domain of word recognition and spelling correction is presented to show the power and versatility of the system.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available