[Study on Identification of Immature Corn Seed Using Hyperspectral Imaging]

Area of Science:

• Agricultural Science
• Spectroscopy
• Image Processing

Background:

• Seed maturity is crucial for seed vigor and consistency.
• Separating immature seeds during sorting enhances seed lot quality.
• Hyperspectral imaging offers a non-destructive method for seed quality assessment.

Purpose of the Study:

• To identify sensitive spectral bands for corn seed maturity detection.
• To develop an image classification method for immature corn seeds.
• To provide a basis for automated seed sorting technology.

Main Methods:

• Hyperspectral imaging (400-1000 nm) and Principal Component Analysis (PCA).
• Partial Least Squares Regression (PLSR) for spectral analysis.
• Band ratio algorithm (640 nm/525 nm) and Kruskal-Wallis test for optimal band selection.

Main Results:

• PC2 from PCA highlighted spectral differences between mature and immature seeds.
• A band ratio of 640 nm/525 nm effectively distinguished seed maturity.
• The band ratio method achieved a 93.9% correct recognition rate for immature corn seeds.

Conclusions:

• Hyperspectral imaging, particularly using the 640 nm/525 nm band ratio, accurately differentiates immature corn seeds.
• This technique enables effective separation of mature and immature seeds based on area ratio.
• The findings support the development of advanced automated seed sorting devices.