Defect Classification of Electronic Board Using Multiple Classifiers and Grid Search of SVM Parameters

Abstract

This paper proposes a new method to improve the classification accuracy by multiple classes classification using multiple SVM. The proposed approach classifies the true and pseudo defects by adding features to decrease the incorrect classification. This approach consists of two steps. First, the features are extracted from the defect candidate region after extracting the difference between the test image and the reference image. Here, candidate extraction is carefully extracted with high accuracy and the useful combination of features is determined using the feature selection. Second, selected features are learned with multiple SVM and classified into the class. When the result has the multiple same voting counts to the same class, the judgment is treated as the difficult class for the classification. It is shown that the proposed approach gives efficient classification with the higher classification accuracy than the previous approaches through the real experiment.