Defect Classification Algorithm for IC Photomask Based on PCA and SVM

Abstract

During IC photomask vision inspection, considering problem that fine image defectpsilas fineness, complex shape, extraction feature difficultly, and effect by noise easily, presented defect identification classification algorithm based on PCA (principal components analysis) and SVM (support vector machine). It resolved the problem that fine and complex defect was difficult to classify, by merits of the extracting image global feature with PCA, and high accuracy and generalization capability with SVM. Regard class distance as criterion to construct the binary tree in multi-class SVM classification algorithm. It resolved the problem that the structure of binary tree affected the accuracy of classifier, and upgraded defect classification accuracy finally. Experiments show that six defects classification accuracy by this method is up to 97.8%, higher than best accuracy 93.3% by BP network and 83.3% by method based on region. And the training and inspecting time is few. In result, itpsilas an effective method for fineness defect identification and classification.