New Analytical Strategies for Quality Control and Classification of Apple Juices Using Digital Image Processing (DIP)...

Area of Science:

• Food Science
• Analytical Chemistry
• Computer Science

Background:

• Apple juice adulteration is a significant concern due to its commercial value.
• Existing analytical methods can be costly, time-consuming, or invasive.
• There is a need for rapid, non-destructive techniques to ensure apple juice authenticity and quality.

Purpose of the Study:

• To develop and validate a smartphone-based image analysis system for apple juice authentication.
• To classify different types of apple-based beverages (whole juice, reconstituted juice, nectar).
• To predict the actual percentage of apple juice content in commercial beverages.

Main Methods:

• Collected images of various apple juice samples (whole juice, reconstituted, nectar).
• Employed machine learning algorithms, including k-nearest neighbors (kNN) and extreme gradient boosting (XGBoost), for classification and prediction.
• Developed calibration curves and utilized cross-validation for model accuracy assessment.

Main Results:

• Classification models achieved high accuracies (e.g., 95.9% in testing for kNN).
• Predictive models demonstrated strong performance with high coefficients of determination (R² ≈ 93.1-93.5% in testing).
• The approach proved effective in distinguishing beverage categories and estimating juice content.

Conclusions:

• Smartphone imaging coupled with machine learning provides a scientifically novel and practical solution for apple juice analysis.
• This method is rapid, accurate, non-destructive, and cost-effective.
• The technology holds significant potential for industrial quality control and regulatory applications.