The Classification of ITO Glass Defects by Image Processing and Neural Network
| Autor: | Chao-chih Chen, 陳朝治 |
|---|---|
| Rok vydání: | 2007 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 In this study, the defects of Indium Tin-Oxide glass ( ITO Glass ) are classified by using image processing and neural networks, The common defects of an ITO Glass include crack, scratch, dust and other blemishes. From the scanned images of defected ITO glasses, we calculated four defect features based on the texture and proportion of geometry characteristics of the defected area. Four defect features include the ratio of defected circumference to aspect circumference, aspect ratio, the density of the defected area and the ratio of defected circumference to defected area. The used image processing technologies include median filter, binary image and edge detection. A three-layer neural network, which is trained by the back-propagation algorithm, is used as a classifier with these four input features. The neural network is trained by forty training samples. The classification accuracy for four defect types by using another twenty testing samples is about 95%. The average error of four classification outputs is 1.82%. The experimental results show that the proposed approach could be effectively applied to inspect ITO glass defects to reduce manual classification errors. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|