Automatic Classification of Textile Fibers and Content Detection Based on Image Processing

In order to improve the identification efficiency of textile fibers and obtain the blending ratio of textile fibers, it is urgent to have an accurate and rapid fiber identification method. In this paper, four new fiber features are proposed in terms of fiber feature extraction: average 2D Gabor bidirectional line features, curvature, fitted curvature and segmented chord length angle change. The average diameter, rectangularity, grayscale symbiotic matrix, average 2D Gabor bipolar line, curvature, fitted curvature, and segmented chord length angle variation were extracted for fibers in total 7 Species characteristics are normalized, using decision tree SVM for flax fibers, modal cotton and viscose fibers Cotton is classified and compared with different nuclear functions, and the data shows that the classification accuracy of the Linear nuclear function is better, and the recognition rate of Modal cotton, linen cotton, and viscose cotton can reach 93.4%, 88.6%, 87.7%. Finally, the average diameter-number method is used to complete the fiber blending ratio test.