Maize kernel hardness classification by near infrared (NIR) hyperspectral imaging and multivariate data analysis.

Area of Science:

• Agricultural Science
• Spectroscopy
• Image Analysis

Background:

• Maize kernel texture (hard, intermediate, soft) is crucial for processing and quality.
• Objective classification of maize kernel texture is needed for breeding and industry.

Purpose of the Study:

• To evaluate near-infrared (NIR) hyperspectral imaging for distinguishing maize kernel hardness.
• To assess the effectiveness of hyperspectral image analysis techniques for this classification task.

Main Methods:

• Acquired NIR hyperspectral images of maize kernels using MatrixNIR and SWIR systems.
• Applied principal component analysis (PCA) for image cleaning and feature extraction.
• Utilized partial least squares discriminant analysis (PLS-DA) to build classification models.

Main Results:

• PCA effectively identified histological classes (glassy/hard vs. floury/soft) in maize kernels.
• PLS-DA models achieved low root mean square error of prediction (RMSEP) values (0.18 for MatrixNIR, 0.29 for SWIR).
• Reproducible results across different datasets indicate method robustness.

Conclusions:

• NIR hyperspectral imaging combined with PCA and PLS-DA is a viable method for classifying maize kernel hardness.
• This technique holds significant potential for future applications in maize breeding and quality control.