Automated visual inspection for surface appearance defects of varistors using an adaptive neuro-fuzzy inference system

被引：10

作者：

Su, J. C. ^{[1
]}

Tarng, Y. S. ^{[1
]}

机构：

[1] Natl Taiwan Univ Sci & Technol, Dept Mech Engn, Taipei 106, Taiwan

来源：

INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY | 2008年 / 35卷 / 7-8期

关键词：

automated visual inspection; machine vision; ANFIS; edge detection; classification;

D O I：

10.1007/s00170-006-0756-3

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

The purpose of this study is to develop an automated visual inspection system for analysis of the surface appearance of ring varistors based on an adaptive neuro-fuzzy inference system (ANFIS). Known image patterns of the six types of ring varistors are used in a training process to establish Sugeno FIS rules, and the input-output data are then set to train the ANFIS to tune the membership function. Feature extraction reduces image complexity using two-dimensional edge detection, calculated within divided rectangular region. The ANFIS combines the neural network adaptive capabilities and fuzzy logic qualitative to train a classification system for six different types of components. The performance of the ANFIS is evaluated in terms of training performance and classification accuracy. The results confirm that the proposed ANFIS is capable of classifying the six types of ring varistors with an accuracy of 98.67%.

引用

页码：789 / 802

页数：14