Machine learning classification of mango maturity based on carotene content from Raman spectra.

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Spectroscopy

Background:

• Mango ripeness is crucial for taste, aroma, and nutrition, impacting farmer yields.
• Traditional ripeness assessments are inconsistent, inaccurate, and time-consuming.
• Variations in mango color and human perception affect traditional methods.

Purpose of the Study:

• Develop a non-invasive, efficient method for detecting mango ripeness using Raman spectroscopy.
• Extract organic compound data from raw Raman spectra.
• Correlate carotene characteristics with mango ripeness levels.
• Evaluate machine learning models for mango ripeness classification.

Main Methods:

• Analyzed 29 mango fruit spectra, with 13 samples representing underripe, ripe, and overripe categories.
• Utilized Raman spectroscopy to analyze organic compounds in mango peel.
• Applied statistical analysis and machine learning models (SVM, KNN) for classification.

Main Results:

• Identified carotenoids (lycopene, β-carotene, lutein, neoxanthin) in mango peel (1,480–1,550 cm⁻¹).
• Confirmed significant correlation between carotenoid peak intensity and mango maturity (p < 0.05).
• Achieved 100% accuracy in classifying ripeness using SVM and KNN models.

Conclusions:

• Raman spectroscopy is a reliable, robust, and non-invasive method for assessing mango ripeness.
• The technique is immune to external factors like light, humidity, and noise.
• This approach holds promise for improving mango quality control and harvest management.