Textile fiber component identification method, electronic equipment and storage medium

Abstract

The invention provides a textile fiber component identification method, electronic equipment and a storage medium. The method comprises the steps of obtaining a to-be-identified fiber image; inputting the to-be-identified fiber image into a pre-trained fiber classification model to obtain a first output feature vector of a feature extraction module in the fiber classification model; wherein the fiber classification model is obtained by training a softmax loss function, a triple loss function and a center loss function; and determining the fiber category of the to-be-identified fiber image according to the first output feature vector and the chemical fiber mean value and the cotton and linen fiber mean value corresponding to the fiber classification model. According to the embodiment of the invention, the softmax loss function, the triple loss function and the center loss function are utilized to train the fiber classification model, so that the identifiable features of the cotton and linen fibers and the chemical fibers are found in the training process, and the classification accuracy of the chemical fibers and the cotton and linen fibers is effectively improved.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular, to a method for identifying textile fiber components, electronic equipment and storage media. Background technique [0002] In the apparel industry, in order to facilitate production management and product analysis, it is necessary to scientifically identify the composition of textile fibers. [0003] Among the various fibers of clothing, natural fibers such as cotton, linen and ramie account for a relatively high proportion. In the quantitative and qualitative analysis of cotton and hemp components, the traditional method is to place cotton and hemp natural fibers on glass slides, wash and disperse them, and then use human eyes to statistically classify the fibers in the glass slides under a microscope. The analysis method is time-consuming and labor-intensive; the existing fiber composition analysis method is based on the electron microscope, auto-focusing and...

AI Technical Summary

Problems solved by technology

In the quantitative and qualitative analysis of cotton and hemp components, the traditional method is to place cotton and hemp natural fibers on glass slides, wash and disperse them, and then use human eyes to statistically classify the fibers in the glass slides under a microscope. The analysis method is time-consuming and labor-intensive; the existing fiber composition analysis method is based on the electron microscope, auto-focusing and shooting each part of the glass slide, and then using instance segmentation to obtain a screenshot of each fiber, and then performing quantitative and qualitative analysis on each fiber screenshot. The problem is that there are often a certain amount of chemical fibers (such as polyester, acrylic, nylon) in the cotton and linen fibers of garments. When directly identifying the fiber components of the fibers, the chemical fibers are often misidentified as cotton or linen.

Furthermore, under the electron microscope, the imaging pixels of a single fiber are large, and the texture of the cotton and linen fiber is distributed in the whole single fiber image, which leads to low accuracy of quantitative and qualitative analysis of the cotton and linen fiber

Images