Decision support in machine vision system for monitoring of TFT-LCD glass substrates manufacturing

•Making an automatic inspection system for TFT-LCD glass substrates manufacturing.•Using wavelet co-occurrence signature from substrate images for feature extraction.•Comparing the performance of CART, optimized SVM and MLP classifiers using SA as the best classifier for proposed automatic inspectio...

Full description

Saved in:
Bibliographic Details
Published in:Journal of process control Vol. 24; no. 6; pp. 1015 - 1023
Main Authors: Yousefian-Jazi, Ali, Ryu, Jun-Hyung, Yoon, Seongkyu, Liu, J. Jay
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.06.2014
Subjects:
Automatic optical inspection system
Imbalanced data
Parameter optimization
Supervised classification
Synthetic Minority Over-sampling TEchnique (SMOTE)
TFT-LCD glass substrate
Synthetic Minority Over-sampling TEchnique (SMOTE)
Imbalanced data
TFT-LCD glass substrate
Parameter optimization
Supervised classification
Automatic optical inspection system
ISSN:0959-1524, 1873-2771
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•Making an automatic inspection system for TFT-LCD glass substrates manufacturing.•Using wavelet co-occurrence signature from substrate images for feature extraction.•Comparing the performance of CART, optimized SVM and MLP classifiers using SA as the best classifier for proposed automatic inspection system.•The proposed SVM model turned out to be appropriate in the context of TFT-LCD glass substrates inspection system. This study addresses classification methodology for the automatic inspection of a range of defects on the surface of glass substrates in thin film transistor liquid crystal display glass substrate manufacturing. The proposed methodology consisted of four stages: (1) feature extraction by calculating the wavelet co-occurrence signature from the substrate images, (2) handling of imbalanced dataset using the Synthetic Minority Over-sampling TEchnique (SMOTE), (3) reduction of the feature's dimension by principal component analysis, and (4) finally choosing the best classifier between three different methods: Classification And Regression Tree (CART), Multi-Layer Perceptron (MLP) and Support Vector Machine (SVM). In training the SVM and MLP classifiers, the simulated annealing algorithm was used to obtain the optimal tuning parameters for the classifiers. From the industrial case study, the proposed feature extraction algorithm could remove the defect-irrelevant image features and SMOTE increased the accuracy of all three methods. Furthermore, the optimized SVM and MLP models were more accurate than the CART model whereas a higher accuracy of 89.5% was observed for the proposed SVM model.
ISSN:0959-1524
1873-2771
DOI:10.1016/j.jprocont.2013.12.009