Online testing of fabric flaws based on machine vision
Author(s): YU Zu-yao, WANG Hong-yuan, ZHANG Ji, School of Information and Mathematics, Changzhou University
Pages: 2851-2856
Year: 2016 Issue: 10
Journal: Computer Engineering and Design
Keyword: patterned fabric; machine vision; elementary cycle; flaw testing; classifiers;
Abstract: The machine vision testing method was put forward for the purpose of achieving real-time online testing of the fabric flaws.In off-line training,the superposition distance function and the weight analysis of extremum of the pictures of flawless fabric were used to get the precise elementary cycle of the fabric texture pattern.With the obtained standard elements,the offset sequence for flawless standard elements was established and the fuzzy classifier was formed.In on-line testing,the image block of the untested fabric was captured with the established offset sequence and classified using the fuzzy classifier.If the fabric was flawed,the precise classifier was formed to continue the classification.Experimental results show that the average time on testing one frame of any cloth with the same texture pattern on-line is 150 ms with above 99% accuracy.It has high real-time performance and low fall-out ratio.
Pages: 2851-2856
Year: 2016 Issue: 10
Journal: Computer Engineering and Design
Keyword: patterned fabric; machine vision; elementary cycle; flaw testing; classifiers;
Abstract: The machine vision testing method was put forward for the purpose of achieving real-time online testing of the fabric flaws.In off-line training,the superposition distance function and the weight analysis of extremum of the pictures of flawless fabric were used to get the precise elementary cycle of the fabric texture pattern.With the obtained standard elements,the offset sequence for flawless standard elements was established and the fuzzy classifier was formed.In on-line testing,the image block of the untested fabric was captured with the established offset sequence and classified using the fuzzy classifier.If the fabric was flawed,the precise classifier was formed to continue the classification.Experimental results show that the average time on testing one frame of any cloth with the same texture pattern on-line is 150 ms with above 99% accuracy.It has high real-time performance and low fall-out ratio.
Related Articles