Image-based classification of Golden Melon ripeness using convolutional neural networks and data augmentation

Accurate fruit maturity assessment is critical for postharvest quality control and optimal harvest scheduling. This study investigates the performance of machine learning and deep learning approaches for classifying Golden Alisha melon maturity using image data. A self-developed dataset comprising 230 labelled images was organized into four maturity stages: 47, 53, 60, and 67 days after planting (DAP). Four classification models were evaluated: Principal Component Analysis combined with Support Vector Machines (PCA-SVM), Principal Component Analysis with Neural Networks (PCA–NN), a Convolutional Neural Network (CNN), and a CNN enhanced with data augmentation. The augmentation strategy included random rotations (≤20°), horizontal and vertical shifts (≤10%), zooming (≤20%), and horizontal flipping applied during training. Performance was assessed using precision, recall, F1-score, and overall accuracy. Experimental results demonstrate that the augmented CNN achieved the highest accuracy of 86%, outperforming both PCA-based models and the baseline CNN. Confusion matrix analysis reveals that the 60 DAP stage is the most difficult to classify, likely due to visual similarity with adjacent ripening stages. This study provides a curated image dataset and confirms the effectiveness of data-augmented deep learning for robust agricultural maturity classification.