Tea cultivar classification and biochemical parameter estimation from hyperspectral imagery obtained by UAV

Area of Science:

• Agricultural Science
• Remote Sensing
• Plant Science

Background:

• Classifying plant species using remote sensing is feasible, but identifying cultivars remains challenging.
• Tea (Camellia sinensis L.) cultivars exhibit significant variations in taste and quality.
• Hyperspectral remote sensing offers potential for fine-scale plant classification and biochemical component estimation.

Purpose of the Study:

• To investigate the feasibility of classifying tea (Camellia sinensis L.) cultivars using hyperspectral remote sensing.
• To estimate taste-related biochemical components (tea polyphenols and amino acids) in tea canopies.
• To assess the potential of Unmanned Aerial Vehicle (UAV)-based hyperspectral data for tea quality assessment.

Main Methods:

• Collected hyperspectral data of tea canopies using a UAV-mounted hyperspectral camera.
• Classified tea cultivars based on spectral characteristics using Support Vector Machine (SVM).
• Predicted tea polyphenols (TP) and amino acids (AA) using Partial Least Squares Regression (PLSR) with spectral normalization.

Main Results:

• Tea cultivar classification achieved an overall accuracy exceeding 95% with appropriate spectral pre-processing.
• PLSR models accurately predicted TP and AA content, with the TP to AA ratio showing the highest accuracy (R_CV = 0.66).
• UAV-based hyperspectral remote sensing demonstrated success in differentiating tea cultivars and holds potential for mapping taste-related components.

Conclusions:

• Hyperspectral remote sensing from UAVs is effective for classifying tea (Camellia sinensis L.) cultivars.
• This technology shows promise for non-destructively assessing and mapping key taste-related biochemical components in tea plantations.
• Further research is necessary to enhance the accuracy of biochemical component prediction for practical applications.