Miniaturized NIRS Coupled with Machine Learning Algorithm for Noninvasively Quantifying Gluten Quality in Wheat...

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Spectroscopy

Background:

• Accurate assessment of gluten content (dry gluten content, wet gluten content, gluten index) is crucial for wheat flour quality.
• Traditional methods for gluten analysis are often destructive and time-consuming.
• Non-invasive techniques are needed for rapid and efficient quality control in the food industry.

Purpose of the Study:

• To develop and validate a non-invasive method for quantitative evaluation of dry gluten content (DGC), wet gluten content (WGC), and gluten index (GI) in wheat flour.
• To integrate a miniaturized near-infrared spectroscopy (NIRS) system with machine learning algorithms for gluten analysis.
• To optimize wavelength selection and machine learning models for improved prediction accuracy.

Main Methods:

• Utilized a miniaturized near-infrared spectroscopy (NIRS) system operating in the 900-1700 nm range.
• Employed five different machine learning algorithms, including support vector regression (SVR), to model the relationship between spectral data and gluten parameters.
• Applied five wavelength selection techniques and developed optimized SVR models, including an improved whale optimization algorithm-based SVR (iWOA-SVR).

Main Results:

• Support vector regression (SVR) demonstrated excellent prediction performance for all gluten parameters using full-range spectra (Rₚ = 0.9370-0.9430).
• Optimized models using 25-30 selected wavelengths showed comparable accuracy.
• The iWOA-SVR model achieved the strongest predictive capability among optimized models (Rₚ = 0.9190-0.9385), confirming robustness through external validation.

Conclusions:

• Micro-NIRS combined with iWOA-SVR provides an effective, non-destructive method for assessing wheat flour gluten quality.
• This approach offers a valuable alternative to traditional methods, enabling faster and more efficient quality control.
• The findings support the expansion of NIRS technology for developing portable, specialized equipment for industrial applications.