Development of Pumpkin Ripeness Classification System by Integrating Image and Sound Data Using Machine Learning Techniques

Abstract

Pumpkin is a high-demand economic crop, but determining its ripeness remains a challenge due to limited expertise among farmers and the labor-intensive manual sorting process. This study aimed to develop a classification model for assessing the ripeness levels of "Ampan 342" pumpkins using Brix values from 30 samples categorized into three levels: unripe, ripe, and overripe. Image and sound data were collected and used to train Convolutional Neural Network (CNN) models. The image-based model achieved an F1-score of 0.84 and a sensitivity of 0.85, while the sound-based model yielded an F1-score of 0.87 and a sensitivity of 0.85. The combined image and spectrogram-based model achieved the best performance, with an F1-score of 0.93 and a sensitivity of 0.92. The dataset comprised 900 images (150 for testing), 900 audio recordings (150 for testing), and 1,800 combined image-spectrogram samples (300 for testing). Limitations of the study include the relatively small dataset and the susceptibility of audio data to environmental noise. Nevertheless, the proposed approach shows promise for real-world applications as an automated system for evaluating pumpkin ripeness in the field, enhancing accuracy and reducing post-harvest losses.