Watermelon Classification using Machine Learning with Enhanced Features

The identification of watermelon sweetness using machine learning techniques has become an essential approach to enhance agricultural practices and improve consumer satisfaction. This study explores the potential of utilizing both image and acoustic features for classifying watermelon sweetness using two widely recognized algorithms: Naive Bayes and Decision Tree. Images of watermelons were captured and processed to extract visual features such as entropy, contrast, and color, while acoustic signals produced by tapping the watermelons were analyzed for resonance patterns that correlate with their internal quality. The acoustic features used for watermelon sweetness classification include the mean of Mel-Frequency Cepstral Coefficients (MFCC), spectral centroid, maximum frequency, and magnitude, which capture key characteristics of the sound produced when tapping the watermelon. The extracted data was then used to train and test both classifiers. The performance of the Naive Bayes and Decision Tree models was evaluated using standard metrics, such as accuracy, precision, recall, and F1-score, to determine their effectiveness in predicting the sweetness of watermelons. Results demonstrated that integrating multi-modal features, specifically image and acoustic data, provides a robust method for accurately assessing watermelon quality, with the Decision Tree model showing superior performance. This approach presents a novel and efficient technique for the automated grading of watermelon sweetness, contributing to advancements in food quality assessment and smart agriculture.