LAUSR

Abstract The identification of wool and cashmere is always considered as a challenge in the field of textile and clothing industry. A novel digital analysis method based on fractal algorithm, parallel-line algorithm, and K-mean clustering algorithm was proposed in the paper to improve the accuracy of fiber measurement and identification. First, the original images of 600 cashmere fibers and 600 wool fibers were captured by optical microscope attached with a digital camera, the self-developed image preprocessing was carried out to obtain the binary image and contour image of cashmere and wool fibers. Then, the fractal algorithm was used to calculate the box-counting and information dimension of fiber binary images, and the parallel-line algorithm was used to measure the fiber fineness of the contour image. These features can be used as the representation of fiber surface morphological features. Finally, the cluster analysis of the extracted feature set was carried out by k-means algorithm, the specimens used for the identification include three sets of wool and cashmere fibers blended with different proportions. The experimental results showed that the accuracy of the fractal-based and diameter-based identification of three sets with different fiber blending proportions was higher than that of traditional methods, the average identification accuracy rate was 97.47%. Through this investigation, the novel digital analysis method proposed in the paper was proved to be feasible to measure and identify the wool and cashmere fibers, it had the protentional application of digital and automatic fiber classification in the future, in terms of both equipment, method and standard.