Welcome to the InfoSci Platform
Reference Hub4
Reduction of Defect Misclassification of Electronic Board Using Multiple SVM Classifiers

Reduction of Defect Misclassification of Electronic Board Using Multiple SVM Classifiers

Takuya Nakagawa, Yuji Iwahori, M. K. Bhuyan
Copyright: © 2014 |Volume: 2 |Issue: 1 |Pages: 12
ISSN: 2166-7160|EISSN: 2166-7179|EISBN13: 9781466656598|DOI: 10.4018/ijsi.2014010103
Cite Article Cite Article

MLA

Nakagawa, Takuya, et al. "Reduction of Defect Misclassification of Electronic Board Using Multiple SVM Classifiers." IJSI vol.2, no.1 2014: pp.25-36. http://doi.org/10.4018/ijsi.2014010103

APA

Nakagawa, T., Iwahori, Y., & Bhuyan, M. K. (2014). Reduction of Defect Misclassification of Electronic Board Using Multiple SVM Classifiers. International Journal of Software Innovation (IJSI), 2(1), 25-36. http://doi.org/10.4018/ijsi.2014010103

Chicago

Nakagawa, Takuya, Yuji Iwahori, and M. K. Bhuyan. "Reduction of Defect Misclassification of Electronic Board Using Multiple SVM Classifiers," International Journal of Software Innovation (IJSI) 2, no.1: 25-36. http://doi.org/10.4018/ijsi.2014010103

Export Reference

Mendeley
Favorite Full-Issue Download

Abstract

This paper proposes a new method to improve the classification accuracy by multiple class 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, detect the straight line by Hough Transform to the inspection image and condition is judged with the gradient. More than 80% of AOI images consist of images with the margin line between base part and lead line part which has the same direction. When detected line directions are almost the same directions, shifted image of inspection image is generated and used as the reference image. In case of different directions of detected lines (this case holds for less than 20% of AOI images), reference image is generated manually. After the reference image is prepared, the difference is taken between the inspection image and reference image. This leads to extract the defect candidate region with high accuracy and features are extracted to judge the defect and foreign material. 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.

Request Access

You do not own this content. Please login to recommend this title to your institution's librarian or purchase it from the IGI Global bookstore.