Date of Award

August 2020

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering (Holcomb Dept. of)

Committee Member

Adam Hoover

Committee Member

Richard Brooks

Committee Member

Richard Groff

Abstract

This thesis considers the problem of variability in the appearance of machine parts while performing automated camera-based inspections in an appliance manu- facturing plant. In an appliance manufacturing plant, machine parts are inspected to find any defects which might have occurred while assembling them. Training the system for the different appearances of these parts is important for detecting defects with high accuracy and precision. Machine parts have a lot of variability in their ap- pearance and it is a difficult and time consuming task to train the inspection system for the same. Previously, a tool (clustering) was developed by our group that could automatically learn the different appearances of machine parts that needed inspec- tion [7]. In this thesis, we perform different experiments using the tool with a goal of training an inspection system prototype for the learned variability. To do our exper- iments, we collected a total of 249,371 images for 9 inspection problems. A manual review of the data identified a range of 2-180 defects per problem. Our inspection system prototype, post training, achieved a range of 83% - 100% defect detection with a range of 0.0-2.9 false alarms per defect for 8 out of the 9 inspection problems. One of inspection problems was not firm in its structure, and in our experiments we could only find 50 % of the defects with a false alarms per defect rate of 5.0. Based on these results, we have designed a decision tree that could be used by engineers for training an inspection system for new inspection problems, for appearance variability.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.