A knowledge-based machine vision system

dc.contributor.authorLu, Ning
dc.date.accessioned2020-02-26T09:15:04Z
dc.date.available2020-02-26T09:15:04Z
dc.date.issued1997
dc.descriptionA thesis submitted to the Faculty of Engineering, University of the Witwatersrand, Johannesburg, in fulfillment of the requirements for the degree of Doctor of Philosophyen_ZA
dc.description.abstractAs a result of the work on this project, two machine vision systems are presented in this thesis. The first system is designed for an industrial application which identifies and locates machined components which have general geometric boundaries. The system consists of a CCD camera, a Pi030 image processing system, and a host computer running under the Linux operating system. The problem of extracting the shape information is treated as a two level image processing procedure, viz. Segmentation and decomposition. Objects are first segmented into generic ‘blobs’ and each blob is further decomposed into a primitive chain which constitutes a series of dominant points and basic primitive segments. The recognition task is accomplished by applying the knowledge base techniques. The shape information is used to form an object description using a specific scheme which is a combination of a semantic network, feature vectors and a production system. The knowledge base is constructed with this description scheme. The inference engine is based on an object verification procedure and a blob verification procedure. The inference results are also used as feedback information to improve the image processing procedure. Experimental results have proved that this system has the advantages of accuracy and reliability. In tests conducted on 19 machined components with different combinations of generic features, the classification system showed a 100% success rate. In addition the system is capable of being easily modified and expanded to cater for parts that are touching and/or overlapping as well as parts in motion on a conveyor. The second machine vision system has been separately designed to carry out the fruit inspection and classification tasks automatically. The system uses a low cost Personal Computer (PC) and an image acquisition hardware to perform these tasks. The concept of fuzzy sets is employed to design the fruit classifier. Preliminary results have proved the feasibility of the system in which is felt to be suitable for individual farmers who wish to carry out automatic fruit classification.en_ZA
dc.description.librarianAndrew Chakane 2020en_ZA
dc.identifier.urihttps://hdl.handle.net/10539/28954
dc.language.isoenen_ZA
dc.subjectComputer vision.en_ZA
dc.subjectArtifical intelligence.en_ZA
dc.subjectImage processing.en_ZA
dc.subjectPattern recognition systems.en_ZA
dc.subjectRobot visions.en_ZA
dc.titleA knowledge-based machine vision systemen_ZA
Files
Original bundle
Now showing 1 - 2 of 2
No Thumbnail Available
Name:
Lu Ning._A knowledge-based machine vision system_001.pdf
Size:
4.3 MB
Format:
Adobe Portable Document Format
Description:
No Thumbnail Available
Name:
Lu Ning._A knowledge-based machine vision system_002.pdf
Size:
950.18 KB
Format:
Adobe Portable Document Format
Description:
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Collections