Rapid Evaluation of Wet Gluten Content in Wheat Using Hyperspectral Technology Combined with Machine Learning...

Area of Science:

• Agricultural Science
• Spectroscopy
• Machine Learning

Background:

• Accurate and rapid wheat quality evaluation is crucial for the food industry.
• Traditional methods for assessing wheat quality, such as wet gluten content, are often time-consuming and labor-intensive.
• Hyperspectral imaging offers a promising non-destructive technique for analyzing agricultural products.

Purpose of the Study:

• To investigate the performance of various machine learning algorithms for predicting wheat wet gluten content.
• To optimize hyperspectral data preprocessing techniques for improved prediction accuracy.
• To establish a rapid and intelligent method for wheat quality assessment.

Main Methods:

• Collected visible and near-infrared (Vis-NIR) hyperspectral data from wheat grains and flour.
• Applied and compared several machine learning algorithms, focusing on random forest regression (RFR).
• Optimized spectral data using first-derivative (FD) transformation and Savitzky-Golay (SG) filtering.

Main Results:

• Random Forest Regression (RFR) demonstrated superior predictive performance for wet gluten content.
• Optimized visible spectra (SG-filtered FD) achieved high prediction accuracy (r²=0.8579 for grains, r²=0.8383 for flour).
• Fused Vis-NIR data also showed strong predictive power when processed with RFR and optimization techniques.

Conclusions:

• The optimized hyperspectral imaging combined with RFR provides an efficient and accurate method for wheat quality evaluation.
• This intelligent prediction scheme has significant potential for real-time quality control in food processing.
• The study highlights the effectiveness of machine learning in agricultural product analysis.