Identification of wool and cashmere SEM images based on SURF features. (July 2019)

• Record Type: Journal Article
• Authors: Lu, Kai; Luo, Junli; Zhong, Yueqi; Chai, Xinyu
• Is Part Of: Journal of engineered fibers and fabrics. Volume 14(2019)
• Publication Date: 2019-07
• Subjects: Wool -- cashmere -- image -- SURF -- feature -- identification; Nonwoven fabrics -- Periodicals; Fibers -- Periodicals; Fibers; Nonwoven fabrics; Periodicals; 677.6
• DOI: 10.1177/1558925019866121
• Languages: English
• ISSNs: 1558-9250

Abstract

Pattern recognition and feature extraction methods are applied to identify cashmere and wool fibers, which are two kinds of very similar animal fibers. In this article, we proposed a new identification method based on Speed Up Robust Features of fiber images. The images of wool and cashmere fibers are obtained by scanning electron microscopy. Speed Up Robust Features of fiber images are extracted, and each fiber image is regarded as a collection of feature vectors in our logic. The vectors are fed into a support vector machine for supervised learning. The findings from scanning electron microscope images indicate that this method is effective; the recognition rate is higher than 93% for a broad range of blend proportions of the two fibers.