Qualitative and Quantitative Assessments of Apple Quality Using Vis Spectroscopy Combined with Improved...

Area of Science:

• Agricultural Science
• Spectroscopy
• Machine Learning

Background:

• Accurate fruit quality assessment is crucial for market value.
• Soluble Solid Content (SSC) is a key indicator of apple quality.
• Visible (Vis) spectroscopy offers a non-destructive method for quality evaluation.

Purpose of the Study:

• To develop a cost-effective and highly accurate optical detection method for apple quality assessment.
• To establish a model for both qualitative and quantitative evaluation of apple SSC using Vis spectroscopy.
• To enhance spectral data and optimize machine learning models for improved accuracy and speed.

Main Methods:

• Visible (Vis) spectroscopy was employed for apple quality assessment.
• Six spectral pretreatment methods, including second-order derivative (SD) and Savitzky-Golay (SG) smoothing, were applied.
• Principal Component Analysis (PCA) was used for spectral enhancement.
• A Back-Propagation Neural Network (BPNN) was utilized for classification, optimized with a Gaussian dynamic learning rate nonlinear decay (DLRND) strategy and particle swarm optimization (PSO).

Main Results:

• The optimized SD-SG-PCA-PSO-BPNN model achieved 100% classification accuracy for apple SSC.
• Quantitative assessment showed a high correlation coefficient (r=0.998) and low root-square-mean error for prediction (RMSEP=0.112 °Brix).
• The developed method demonstrated superior performance compared to a commercial fructose meter.

Conclusions:

• Visible spectroscopy combined with advanced data processing and machine learning offers a powerful tool for apple quality evaluation.
• The proposed synthetic model provides significant value for both qualitative and quantitative assessments of apple SSC.
• This approach holds promise for cost-effective and high-accuracy fruit quality grading systems.