During the salmon can-filling process, a number of can-filling defects can result from the incorrect
insertion of the salmon meat into the cans. These can-filling defects must be repaired before
sealing the cans. Thus, in the existing industrial process, every can is manually inspected to
identify the defective cans. This thesis details a research project on the use of machine vision
for the inspection of filled cans of salmon. The types of can-filling defects were identified and
defined through consultations with salmon canning quality assurance experts. Images of can-filling
defects were acquired at a production facility. These images were examined and feature extraction
algorithms were developed to extract the features necessary for the identification of two types
of can-filling defects. Radial basis function networks and fuzzy logic methods for classifying the
extracted features were developed. These classification methods are evaluated and compared. A
research prototype was designed and constructed to evaluate the machine vision algorithms on-line.
Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:BVAU.2429/7749 |
Date | 05 1900 |
Creators | O’Dor, Matthew Arnold |
Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
Language | English |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Relation | UBC Retrospective Theses Digitization Project [http://www.library.ubc.ca/archives/retro_theses/] |
Page generated in 0.0015 seconds