Scholarlink

<p>This study proposes a chemical analysis-based framework for wood species discrimination by integrating Fourier-transform infrared spectroscopy (FT-IR), gas chromatography–mass spectrometry (GC–MS), and chemometric analysis. After data standardization and principal component analysis (PCA) dimensionality reduction, the leading principal components were used to construct classification models. Several multivariate approaches, including <i>k</i>-nearest neighbors (KNN), linear discriminant analysis (LDA), and artificial neural networks (ANN), were evaluated to examine classification behavior under limited sample conditions. While ANN and KNN showed reasonable classification performance, ANN results varied across repeated trials, reflecting sensitivity to small datasets. In contrast, LDA provided more stable and interpretable discrimination, particularly when FT-IR and GC–MS data were combined. The combined LDA model consistently outperformed single-technique models in cross-validation, demonstrating the complementary value of structural information captured by FT-IR and extractive-related chemical features revealed by GC–MS. The influence of FT-IR spectral range selection was also examined, and restricting analysis to chemically relevant regions improved model stability as the number of species increased. Overall, the results demonstrate that integrating multiple chemical descriptors enables effective species-level discrimination of hardwoods under practical sampling constraints and supports future expansion to broader taxonomic datasets.</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

Scalable wood species classification by integrated chemometrics analysis using Fourier-transform infrared spectroscopy and gas chromatography–mass spectrometry

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Haiquan Chen,
  • Dan Aoki,
  • Te Ma,
  • Akira Watanabe,
  • Tomoya Imai,
  • Kazuhiko Fukushima

摘要

This study proposes a chemical analysis-based framework for wood species discrimination by integrating Fourier-transform infrared spectroscopy (FT-IR), gas chromatography–mass spectrometry (GC–MS), and chemometric analysis. After data standardization and principal component analysis (PCA) dimensionality reduction, the leading principal components were used to construct classification models. Several multivariate approaches, including k-nearest neighbors (KNN), linear discriminant analysis (LDA), and artificial neural networks (ANN), were evaluated to examine classification behavior under limited sample conditions. While ANN and KNN showed reasonable classification performance, ANN results varied across repeated trials, reflecting sensitivity to small datasets. In contrast, LDA provided more stable and interpretable discrimination, particularly when FT-IR and GC–MS data were combined. The combined LDA model consistently outperformed single-technique models in cross-validation, demonstrating the complementary value of structural information captured by FT-IR and extractive-related chemical features revealed by GC–MS. The influence of FT-IR spectral range selection was also examined, and restricting analysis to chemically relevant regions improved model stability as the number of species increased. Overall, the results demonstrate that integrating multiple chemical descriptors enables effective species-level discrimination of hardwoods under practical sampling constraints and supports future expansion to broader taxonomic datasets.