Vision Based For Classification Of MIG Butt Welding Joint Defect Using Occurrence Matrices And Gray Absolute Histogram

Mohd Shah, Hairol Nizam and Sulaiman, Marizan and Shukor, Ahmad Zaki and Kamis, Zalina and Ab Rashid, Mohd Zamzuri (2017) Vision Based For Classification Of MIG Butt Welding Joint Defect Using Occurrence Matrices And Gray Absolute Histogram. International Journal Of Mechanical & Mechatronic Engineering (Ijmme), 17 (5). pp. 60-66. ISSN 2227-2771

Text IJMME2.pdf - Published Version Download (463kB)

Abstract

This paper will introduced a new approach of vision which is enable to overcome the problems in the vision inspection systems. This system uses 2D gray pixels coocurrence matrix and gray absolute histogram of edge amplitude as the input features extract from the MIG butt welding joints. Images of the welding surfaces are captured using one CCD camera that is mounted on the top which is parallel with the work benches. The images are segmented and the 2D gray value coocurrence matrix consists of energy, correlation, homogeneity and contract, and absolute histogram of the characteristic feature in these images will be calculated. The same process will be applied in zooming image factor by 0.5 to calculated the next characteristic feature values. Finally both feature value is used as the input value in GMM and MLP classifier to classify the welds defect into three categories which are good weld, excess weld and insufficient weld. Results are taken from the 18 MIG butt welding joints samples were tested in overall accuracy recognition rate for MLP is 94.4 % while for GMM is 83.3%. In terms of total computation time, the overall time for MLP is 1.96 m/s and GMM is 1.175 m/s.

Item Type:	Article
Uncontrolled Keywords:	Occurrence Matrices, Gray Absolute Histogram, Multi-Layer Perceptrons (MLP), Gaussian Mixture Model (GMM), MIG Butt Welding Joint Defect
Subjects:	T Technology > T Technology (General) T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions:	Faculty of Electrical Engineering > Department of Mechatronics Engineering
Depositing User:	Mohd Hannif Jamaludin
Date Deposited:	27 Jun 2018 08:02
Last Modified:	09 Jul 2021 13:59
URI:	http://eprints.utem.edu.my/id/eprint/20878
Statistic Details:	View Download Statistic

Actions (login required)

View Item