Blending ratio detection of wool/viscose blended fabrics based on deep learning
Aiming at the problem of low efficiency of conventional fabric blending ratio detection methods,a blending ratio detection method of wool/viscose blended fabric based on deep learning technique was proposed.Based on the single-stage target detection algorithm(YOLOv5),optical microscope images of wool fibers and viscose fibers were collected to construct a dataset,fiber features were extracted from the dataset using the CSPDarknet53 network,and the fusion of different levels of features was accomplished by combining FPN and PAN;a convolutional block attention module(CBAM)was introduced in the backbone network to enhance the extraction capability of local features.The average mean precision of the trained YOLOv5 model reached 0.93,which can realize the automatic detection of blending ratio of wool/viscose blended fabrics.The reliability of the model was verified using optical microscopy method and chemical dissolution method,and the results were within 2%difference,indicating that the method has good application prospects in the field of rapid detection of blending ratio of wool/viscose blended textiles.
NETL
NSTL
万方数据
维普
