<![CDATA[The advancement of precision agriculture requires intelligent and automated solutions to enhance crop monitoring and post-harvest management. Accurate recognition of tomato ripeness is a critical stage in the agricultural supply chain, as misclassification can lead to quality degradation and significant economic losses. This study proposes a deep learning-based approach for tomato ripeness detection by integrating handcrafted color features and Gray Level Co-occurrence Matrix (GLCM)-based texture features with YOLOv8. The dataset comprises tomato images categorized into three final ripeness classes—green, semi-ripe, and fully ripe—obtained by grouping multiple physiological maturity stages into unified model classes to reduce visual ambiguity and collected under diverse lighting and background conditions. Candidate tomato regions are first localized by YOLOv8; HSV and CIELAB color descriptors and GLCM texture descriptors are then extracted from each region and fused with the corresponding deep representation for ripeness classification and confidence refinement. Experimental results based on five-fold cross-validation show that the proposed method achieves a mean Average Precision (mAP@0.5) of 96.7%, compared with 92.5% for standard YOLOv8. The results show improved discrimination between visually similar ripeness stages, particularly under non-uniform illumination. The study demonstrates the potential of domain-informed feature fusion for smart farming, automated sorting, and quality control systems.]]>
Copyrights © 2026
