Improving accuracy of automatic optical inspection with machine learning

Abstract

Electronic devices require the printed circuit board (PCB) to support the whole structure, but the assembly of PCBs suffers from welding problem of the electronic components such as surface mounted devices (SMDs) resistors. The automated optical inspection (AOI) machine, widely used in industrial production, can take the image of PCBs and examine the welding issue. However, the AOI machine could commit false negative errors and dedicated technicians have to be employed to pick out those misjudged PCBs. This paper proposes a machine learning based method to improve the accuracy of AOI.In particular, we propose an adjacent pixel RGB value based method to pre-process the image from the AOI machine and build a customized deep learning model to classify the image. We present a practical scheme including two machine learning procedures to mitigate AOI errors. We conduct experiments with the real dataset from a production line for three months, the experimental results show that our method can reduce the rate of misjudgment from 0.3%-0.5% to 0.02%-0.03%, which is meaningful for thousands of PCBs each containing thousands of electronic components in practice.

Fast plug-in capacitors polarity detection with morphology and SVM fusion method in automatic optical inspection system

“…The approach has an accuracy of 87.2% and a recall of 98.9%. Tong et al [10] proposed an adjacent pixel RGB-based method to preprocess the image and build a customized deep learning model to classify the image. The method claims that it can reduce the misjudgment rate from 0.3 − 0.5% to 0.02 − 0.03%.…”

Research on anti-breast cancer candidate drugs analysis based on random forest model and genetic algorithm

“…Accuracy is a commonly used evaluation criterion. However, when the sample size of different categories is quite different, that is, when the sample number is unbalanced, it may cause a large error in practice [8]. Considering that the analytical data used in this study may exist the problem of unbalancing between positive and negative classification, this study used the AUC and ROC curve to make sure of the stabilization of the evaluation results when the distribution of samples changes [9].…”

“…The mathematical model of goal programming that constructed in this paper is as follows: (8) In the formula above, x i represents 20 molecular characteristic variables extracted after data dimension reduction, and y j represents 5 ADMET properties. The multiplication of x and y should be the largest, with the value range of x from 0 to 20, and the value range of y from 0 to 5.…”