期刊名称：International Journal of Multimedia and Ubiquitous Engineering
印刷版ISSN：1975-0080
出版年度：2016
卷号：11
期号：6
页码：93-106
DOI：10.14257/ijmue.2016.11.6.09
出版社：SERSC
摘要：A new detection approach is proposed to detect various uniform and structured fabric defects based on the multiple Gabor filters and Kernel Principal Component Analysis. First of all, images are filtered by multiple Gabor filters with six scales and four orientations to extract feature vectors. After that, the sub-blocks divided from the feature vectors have been fused and the high-dimension data can be reduced by using Kernel Principal Component Analysis. Finally, the similarity matrix is calculated by Euclidean norm and segmented with OTSU threshold method. The experiment has been done by integrating hardware and NI LabVIEW graphical programming language. Experimental results show that proposed algorithm improves feature extraction capability significantly and has high recognition rate.
Loading...
