NIRS-based fresh grape ripeness prediction with SPA-LASSO spectral feature selection

Area of Science:

• Agricultural Science
• Spectroscopy
• Chemometrics

Background:

• Grape quality and optimal harvest timing are critical for winemaking and fresh consumption.
• Accurate, non-destructive methods for assessing grape maturity are highly desirable.
• Near-Infrared Spectroscopy (NIRS) offers a rapid and non-invasive approach for quality assessment.

Purpose of the Study:

• To develop a rapid, non-destructive maturity evaluation model for grapes using NIRS.
• To monitor quality parameter changes during the ripening process.
• To determine the optimal harvest period for Cabernet Sauvignon grapes.

Main Methods:

• Physicochemical analysis of grape parameters across twelve growth stages.
• Preprocessing of spectral data and feature wavelength selection using SPA-LASSO.
• Development of prediction models for Soluble Solid Content (SSC) and Total Acid (TA) using Partial Least Squares Regression (PLSR).

Main Results:

• Soluble Solid Content (SSC) generally increased, while Total Acid (TA) decreased during ripening.
• The SG + SPA-LASSO + PLSR model demonstrated high accuracy for both SSC (R² > 0.98) and TA (R² > 0.94) prediction.
• SPA-LASSO effectively selected relevant wavelengths, improving model generalization.

Conclusions:

• NIRS combined with SPA-LASSO and PLSR provides an effective tool for non-destructive grape maturity assessment.
• The developed model can accurately predict SSC and TA, aiding in the determination of optimal harvest timing.
• This approach enhances spectroscopic screening for grape quality monitoring.