Decision Support in Machine Vision System for Monitoring of TFT-LCD Glass Substrates Manufacturing with Data Mining Techniques

Abstract

The visual appearance of manufactured products is often one of the major quality attributes for certain types of products, which are used mainly for display purposes or used as the exterior part of other products. TFT-LCD (Thin Film Transistor – Liquid Crystal Display) glass substrates can serve as a representative case. Inline defect inspection plays an important role in production yields quality improvement in TFT-LCD manufacturing. The main objective of this work is presenting one decision support system for monitoring of TFT-LCD glass substrates manufacturing by using data mining techniques. This study employs optical system design to make an inline surface defect inspection system in cold process section which is carried out according to CRISP-DM standard for data mining process methodology. This study also develops an image processing methodology, wavelet co-occurrence signature, to extract the features from images and different statistical, heuristical and machine learning algorithms such as principle component analysis, simulated annealing, support vector machine (SVM), multilayer perceptron (MLP) and ensemble techniques are used as feature reduction and classification as well. Finally, the results of different feature selection methods and classifiers are compared and the best is proposed as a suitable methodology for an automatic inspection system in cold process of TFT-LCD glass substrates manufacturing.