Scholarlink

<p>The polyester content directly determines the recycling value and application of polyester blended fabrics: Blends with high polyester content (≥ 50%) have higher recycling value, while those with low content (&lt; 50%) are relegated to lower-value uses. Therefore, achieving rapid and accurate in-situ quantification of polyester content in textiles is essential for enhancing recycling efficiency and economic returns. This study proposes a method combining hyperspectral imaging (1000–1700&#xa0;nm) with a stratified prediction model to detect the content of polyester blended fabrics. The study proposed a stratified preprocessing approach that classified samples into two categories (polyester content ≥ 50% and &lt; 50%) using a machine learning-based grouping strategy. After partitioning the processed dataset into two subsets, principal component analysis was performed on each subset to extract category-specific spectral features. These spectral features were used as inputs to construct dedicated machine learning-based regression models for polyester content prediction in each respective subset. Experimental results demonstrate that the stratified prediction model achieves a root mean square error of 2.87%, significantly outperforming the conventional prediction model (root mean square error: 8.37%). This study enhances the subsequent recycling and regeneration of polyester blended fabrics while providing a theoretical foundation for the industrial application of polyester content quantification.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

In-situ rapid detection of polyester content blended fabrics using hyperspectral imaging and a stratified prediction model

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • Qiyu Gao,
  • Liqiang Zhang,
  • Fei Wang,
  • Wei Xiong,
  • Haibin Cui,
  • Xinrong Wu,
  • Wenyuan Wang,
  • Lihong Zhang

摘要

The polyester content directly determines the recycling value and application of polyester blended fabrics: Blends with high polyester content (≥ 50%) have higher recycling value, while those with low content (< 50%) are relegated to lower-value uses. Therefore, achieving rapid and accurate in-situ quantification of polyester content in textiles is essential for enhancing recycling efficiency and economic returns. This study proposes a method combining hyperspectral imaging (1000–1700 nm) with a stratified prediction model to detect the content of polyester blended fabrics. The study proposed a stratified preprocessing approach that classified samples into two categories (polyester content ≥ 50% and < 50%) using a machine learning-based grouping strategy. After partitioning the processed dataset into two subsets, principal component analysis was performed on each subset to extract category-specific spectral features. These spectral features were used as inputs to construct dedicated machine learning-based regression models for polyester content prediction in each respective subset. Experimental results demonstrate that the stratified prediction model achieves a root mean square error of 2.87%, significantly outperforming the conventional prediction model (root mean square error: 8.37%). This study enhances the subsequent recycling and regeneration of polyester blended fabrics while providing a theoretical foundation for the industrial application of polyester content quantification.