摘要
废旧纺织品的循环利用是纺织品领域实现绿色发展的关键,而其成分识别及分类是废旧纺织品回收过程中的首要及重要环节。将近红外光谱技术(NIR)与谱图处理方法相结合,分别利用卷积神经网络(CNN)和支持向量机(SVM)方法建立了废旧纺织品定性分析模型。其中,在数据预处理阶段,当采用S-G平滑,CNN在内部训练集上的分类准确率为96.3%。相比之下,SVM在最大最小归一化预处理后获得了98.6%的分类准确率,显示该方法可能更适用于混合纺织品识别。
The recycling of waste textiles is the key to achieving green development in the textile industry,and the identification and classification of their components are the primary and important steps in waste textile recycling.Near-infrared spectroscopy technology and spectral processing methods were combined,and qualitative analysis models for waste textiles using convolutional neural network(CNN)and support vector machine(SVM)methods were established,respectively.During the data preprocessing stage,the CNN achieved a classification accuracy of 96.3%on the internal training set when S-G smoothing was applied.In contrast,the SVM model,preprocessed with Min-Max Normalization,attained a higher training accuracy of 98.6%,suggesting that this approach may be more suitable for the mixed textile identification.
作者
王悦
杜宇君
郑佳辉
李宁宁
桑俊锋
李文霞
Wang Yue;Du Yujun;Zheng Jiahui;Li Ningning;Sang Junfeng;Li Wenxia(School of Materials Design and Engineering,Beijing Institute of Fashion Technology,Beijing 100029,China;China Textile Standard Testing and Certification Co.,Ltd.,Beijing 100029,China;General Technology Advanced Materials Group Co.,Ltd.,Beijing 100029,China;China Fiber Quality Monitoring Center,Beijing 100029,China;China Leather and Footwear Industry Research Institute,Beijing 100029,China)
出处
《合成技术及应用》
2025年第2期27-32,共6页
Synthetic Technology & Application
基金
中国纺织工业联合会“纺织之光”应用基础研究项目(J202204)。
关键词
废旧纺织品
近红外光谱技术
卷积神经网络
支持向量机
waste textiles
near infrared spectroscopy technology
convolutional neural network
support vector machine
