Automatic defect classification for TFT-LCD cell process inspection using wavelet transform reconstruction and fuzzy support vector machine

摘要：

Inline defect inspection plays an important role in yield improvement for TFT-LCD manufacturing. This paper proposes a defect identification system by which the defects can be automatically identified in cell process. The proposed system is composed of four parts: preprocessing defect image, wavelet and inverse wavelet transform, feature extraction, and defect identification and classification. For the defect identification and classification, a novel classifier called fuzzy support vector machine (FSVM) with a radius based membership setting is proposed. FSVM is proposed to solve the critical problem existing in traditional standard SVM, overfitting due to outliers and noises. Experimental results, carried out by real defective images provided by a Taiwan TFT-LCD manufacturer, predict that the proposed FSVM outperforms the standard SVM, in terms of defect classification accuracies.

关键词： defect classification Fuzzy support vector machine (FSVM) Inline defect inspection TFT-LCD Support vector machine (SVM)

作者: Te-Sheng Li

作者单位: Department of Industrial Engineering and Management, Minghsin University of Science and Technology, Hsin-Hsin Rd., Hsinchu, 30401, Taiwan

会议类型: 国际会议

会议名称: 2011 Seventh International Conference on Natural Computation(第七届自然计算国际会议 ICNC 2011)

会议地点: 上海

会议语种:英文

页码: 570-574

在线出版日期: 2011-07-26（万方平台首次上网日期，不代表论文的发表时间）