Nondestructive Detection of Polyphenol Oxidase Activity in Various Plum Cultivars Using Machine Learning and Vis/NIR...

Area of Science:

• Agricultural Science
• Spectroscopy
• Biochemistry

Background:

• Polyphenol oxidase (PPO) causes browning and quality loss in harvested fruit.
• Accurate PPO activity measurement is crucial for maintaining fruit market quality.

Purpose of the Study:

• To develop a non-invasive analytical framework for estimating PPO activity in plums.
• To compare the efficacy of various chemometric models for PPO activity prediction.

Main Methods:

• Visible/Near-Infrared (VIS/NIR) spectroscopy was employed for data acquisition.
• Chemometric modeling, including Support Vector Regression (SVR), Decision Tree (DT), and Partial Least Squares Regression (PLSR), was utilized.
• Metaheuristic feature selection identified key wavelengths for model optimization.

Main Results:

• Non-linear models (DT and SVR) outperformed linear PLSR, indicating complex spectral-PPO relationships.
• The Decision Tree (DT) model demonstrated superior generalization performance.
• Optimized DT models using reduced wavelengths enhanced prediction accuracy and reduced computational cost.

Conclusions:

• VIS/NIR spectroscopy combined with optimized DT modeling offers a robust, rapid, and non-destructive method for PPO activity quantification in plums.
• This approach facilitates field-realistic quality assessment of fruit without physical damage.