based drone pineapple maturity detection : fuzzy logic and yolo approach Aburquez, John Cody S. creator Aratea Jr., Sherwin S. author Bautista, Mark Francis A. author text 2026 monographic eng

print

x, 131 pages ; 28 cm. The manual assessment of pineapple ripeness remains a time-consuming and inconsistent practice among local farmers, often resulting in delayed harvesting and post-harvest losses. This study developed a drone-based system capable of detecting pineapples and classifying their maturity to support precision agriculture and reduce manual labor in harvest decisions. A DJI Mini 3 drone captured aerial footage of pineapple fields, which was processed using a YOLO object detection model to locate pineapples and a Mamdani-type fuzzy logic system to classify ripeness as unripe, ripe, or overripe based on peel color, namely green, yellow, and brown. The YOLO model was trained on 4,622 raw and 7,766 augmented images, attaining a precision of 75%, recall of 74%, mAP50 of 76%, and F1 score of 75%. In laboratory testing, the system correctly classified 12 of 15 samples for an overall accuracy of 80%, with an average confidence score of 85.93%. The fuzzy logic component performed best for the Ripe category, with precision, recall, and F1-score of 1.00, while the Unripe category recorded a lower recall of 0.40 due to inconsistent lighting. Reliability measures, namely Cohen's Kappa (k = 0.700), Fleiss' Kappa (к = 0.691), and Krippendorff's Alpha (a=0.701), indicated substantial agreement with ground truth observations, supporting the system's potential to improve harvesting efficiency in the local pineapple industry. John Cody S. Aburquez, Sherwin S. Aratea Jr., Mark Francis A. Bautista, and Carl Darwin M. Gundran. Experimental-developmental research Agile method Includes bibliographical references T ENG A28 2026 261110 20261110105300.0