Research on moldy tea feature classification based on WKNN algorithm and NIR hyperspectral imaging.

Area of Science:

• Agricultural Science
• Food Science
• Spectroscopy

Background:

• Mold contamination in dried tea poses a significant food safety concern.
• Accurate and rapid detection of moldy tea leaves is crucial for quality control.
• Traditional microbiological methods are time-consuming and destructive.

Purpose of the Study:

• To develop a fast and nondestructive method for identifying moldy tea leaves.
• To select effective characteristic wavelengths using hyperspectral imaging.
• To classify dried tea samples based on their mildew degree.

Main Methods:

• Hyperspectral imaging of 300 dried tea samples with varying mildew levels.
• Preprocessing using Roughness Penalty Smoothing (RPS) algorithm.
• Wavelength selection via Wavelet coupled with k-nearest neighbor (WKNN) using different wavelet basis functions (db4, db6, sym5, sym7) and decomposition levels.
• Classification model development using Linear Discriminant Analysis (LDA).

Main Results:

• The WKNN algorithm effectively identified optimal wavelet decomposition levels and characteristic wavelengths.
• The best performing LDA model, utilizing db4 as the wavelet basis function and a decomposition level of 2, achieved 100% identification in the calibration set and 98.33% in the prediction set.
• Food microbiological examination confirmed that mold count increases with storage time, temperature, and humidity.

Conclusions:

• The combination of WKNN and NIR hyperspectral imaging is effective for wavelength extraction and classification of dried tea with different mildew degrees.
• This method offers a rapid and nondestructive approach for detecting mold contamination in tea.
• The developed model demonstrates high accuracy in identifying moldy tea leaves.