Near-Infrared Spectroscopy Combined with Fuzzy Improved Direct Linear Discriminant Analysis for Nondestructive...

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Biotechnology

Background:

• Chrysanthemum tea variety significantly impacts its quality, efficacy, and function.
• Accurate discrimination of chrysanthemum tea varieties is crucial for the tea industry.
• Current methods for variety identification may not be non-destructive or sufficiently accurate.

Purpose of the Study:

• To develop a novel, non-destructive method for categorizing chrysanthemum tea varieties.
• To extract discriminant features from near-infrared (NIR) spectral data.
• To evaluate the effectiveness of a new feature extraction technique, fuzzy IDLDA (FIDLDA).

Main Methods:

• Collected NIR spectral data from 400 samples across five chrysanthemum tea varieties using a portable NIR spectrometer.
• Applied various spectral data preprocessing techniques and principal component analysis (PCA) for dimensionality reduction.
• Employed linear discriminant analysis (LDA), improved direct linear discriminant analysis (IDLDA), and the proposed FIDLDA for feature extraction.
• Utilized the K-nearest neighbor (KNN) algorithm for classification accuracy assessment.

Main Results:

• The proposed FIDLDA method achieved a high discrimination accuracy of 99.2%.
• IDLDA showed improved accuracy at 94.4%, compared to LDA at 87.2%.
• The combination of NIR spectroscopy and FIDLDA demonstrated significant potential for non-destructive variety discrimination.

Conclusions:

• FIDLDA is a highly effective feature extraction method for discriminating chrysanthemum tea varieties.
• Near-infrared spectroscopy coupled with FIDLDA offers a promising approach for non-destructive quality control in the tea industry.
• This technique can aid in ensuring the authenticity and quality of chrysanthemum tea products.