Identification of overlapped wool/cashmere fibers based on multi-focus image fusion and convolutional neural network

Abstract

ABSTRACT Identification of cashmere and wool fibers has always been one of the essential topics in the field of textile. To solve this problem, this paper proposes anew overlapped fiber recognition method based on multi-focus image fusion and convolutional neural network. Firstly, the images of wool and cashmere fibers were captured by an optical microscope and adigital camera. Secondly, the fiber images with clear features were acquired by the multi-focus image fusion method. Then, image preprocessing operations were used to obtain single fiber images. Finally, the features of overlapped fibers are extracted and classified by the classical convolutional neural network. The results show that when the ratio of the training set to the test set is 8:2, the recognition accuracy of the method proposed in this paper reaches the highest of 98.7%, which can be used to realize the fast and accurate identification of overlapped fibers.