A New Method for Component Inspection Based On Machine Learning
The image analysis methods based on features are widely used in commercial automatic optical inspection systems, but the slow inspection speed and high false alarm rate are two problems difficult to solve. In this paper, a new algorithm based on machine learning for inspection of components on print circuit board is proposed. First, to reduce the inspection time, a defect diagnosis decision tree based on image analysis method is established with Classification and Regression Tree. Then, to resolve imbalance problem, we propose Boundary-SMOTE in which the synthetics samples are generated between the boundary samples and their neighbors. Finally, for each node of the defect diagnosis tree, AdaBoost is used to train the extended data set on several features. Experiments results showed that the defect of wrong component, missing component, lacking solder, surplus solder, pseudo joints, shift, tomb stone, etc can be identified properly by using the proposed algorithm.
Automatic optical inspection Decision tree Boundary-SMOTE CART AdaBoost
Xie Hongwei Zhang Xianmin Kuang Yongcong
School of Mechanical and Automotive Engineering, South China University of Technology Guangzhou, Guangdong Province, China
国际会议
2010 International Conference on Measurement and Control Engineering(2010年IEEE测量与控制工程国际会议 ICMCE2010)
成都
英文
55-60
2010-11-16（万方平台首次上网日期，不代表论文的发表时间）