Non-Destructive Testing of Alfalfa Seed Vigor Based on Multispectral Imaging Technology

Area of Science:

• Agricultural Science
• Plant Science
• Spectroscopy

Background:

• Seed vigor is crucial for evaluating seed quality and crop yield.
• Accurate and rapid seed vigor assessment remains a significant challenge in seed research.
• Multispectral imaging offers a promising, non-destructive, and rapid method for seed quality evaluation.

Purpose of the Study:

• To evaluate the effectiveness of multispectral imaging and multivariate analysis for assessing alfalfa seed vigor.
• To distinguish between alfalfa seeds of varying maturity, age, and vigor levels.
• To predict seed viability and germination percentages non-destructively.

Main Methods:

• Collected multispectral data across 19 wavelengths for alfalfa seeds of different maturities and harvest years.
• Applied five multivariate analysis methods: Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), Random Forest (RF), and Normalized Canonical Discriminant Analysis (nCDA).
• Validated the models for distinguishing seed vigor levels and predicting germination.

Main Results:

• Linear Discriminant Analysis (LDA) achieved high accuracy: 92.9% for different maturities and 97.8% for different harvest years.
• LDA demonstrated average sensitivity, specificity, and precision exceeding 90%.
• Normalized Canonical Discriminant Analysis (nCDA) achieved 93.3% accuracy in identifying non-viable seeds and 95.7% for high-vigor seeds and germination prediction.

Conclusions:

• Multispectral imaging coupled with multivariate analysis provides an accurate and rapid non-destructive method for evaluating alfalfa seed vigor, viability, and germination.
• This approach offers significant potential for improving seed quality testing protocols.
• The study establishes valuable technical methods for rapid, non-destructive seed quality assessment.