Attribute-based Maturity Grading of Mango Fruit by Machine Learning

Abstract

Accurate and reliable fruit grading contributes to India’s economic prosperity. Classification and grading are needed before marketing agricultural products. The color, size, shape, and texture of a fruit, as well as its overall look, frequently characterize its quality. Fruit choosing should not be done solely on appearance. Ripeness is largely judged by internal fruit attributes, such as sweetness and acidity. According to studies, a fruit’s maturity is often appraised by an image that depicts its stage. Observing fruits of the same color at different stages of maturation makes determining ripeness difficult. The studied data, which includes qualities with yellow skin pigmentation at all stages of development, is collected from the mango fruit dataset “Nam Dok Mai Si Tong” to address this issue. This study suggests using indicators like Total Soluble Solids (TSS), Titrable Acidity (TA), and BrimA to define mango fruit development according to ripeness for a certain type. Metrics and ten-fold cross-validation is used to analyze ML models. The decision tree classifier has the highest accuracy (91.9% on average) for both unripe and ripe classifications compared to state-of-the-art, according to trial findings.