Automatic inspection of an LCD light-guide plate based on weighted central moments combined with neural networks

Abstract

Liquid crystal displays (LCDs) have become common display devices. Defects of dot patterns on the light-guide plate of an LCD can cause visual failure. A critical task in LCD manufacturing is to detect the micro-defective dot patterns on the LCD light-guide plate to avoid visual failure. This study proposes the integration of weighted central moments with artificial neural networks (ANNs) for the automatic inspection of the LCD light-guide plate, especially for micro-defective dot patterns. The proposed algorithm first finds weighted central moments with adjustable parameters. Then, the back-propagation ANN classifies the dot patterns based on the shape descriptors of weighted central moments. Finally, the ANN algorithm inspects the LCD light-guide plate using the adjustable weighted central moments. The proposed method can successfully classify different dot patterns of the LCD light-guide plate. This study also compares the results of ANN classification and the Bayes classifier. This comparison shows that the proposed method has a higher recognition rate using weighted central moments combined with ANN classification to inspect the LCD light-guide plate.