Rapid and accurate identification of Dendrobium species using FT-IR, FT-NIR, and data fusion with machine learning
Jiaqi Han, Qiang Hu, Yuanzhong Wang
Abstract
Dendrobium is a medicinal plant with anti-aging and hypoglycemic effects, widely used in daily diets. Due to genetic differences, the bioactive components in Dendrobium vary across species. Accurate species identification is crucial for protecting consumer interests and supporting the sustainable development of the market. This study aims to establish a Dendrobium species identification model using FT-NIR, FTIR spectra, and data fusion techniques combined with 2 unsupervised and 11 supervised machine learning algorithms. The results show that the PLS-DA model based on FTIR achieved 100 % accuracy in both training and test sets. The FT-NIR-based 2DCOS-ResNet model classified samples with 100 % accuracy in all sets. The ResNet model is preferred for its inference speed and generalization ability, providing a potential method for plant species identification and quality monitoring.