UNMANNED AERIAL VEHICLE (UAV)-BASED COMPUTER VISION MODEL FOR REAL-TIME BIRDS DETECTION IN RICE FARM

Rice farming in Nigeria suffers significant losses due to bird damage, necessitating advanced mitigation strategies. This study investigates the integration of computer vision with unmanned aerial vehicles (UAVs) to provide real-time bird detection and deterrence in rice fields. Given the varied agricultural conditions in Nigeria—including different farm sizes, vegetation density, and lighting conditions—the proposed system was designed for adaptability and robustness. Utilizing a dataset of 1,113 bird images captured via UAVs and ground cameras, a YOLOv8 model was trained with rigorous preprocessing and augmentation techniques. The model achieved a precision of 85%, recall of 70%, and mAP@50 of 80%, demonstrating strong detection capabilities. However, performance decreased in densely occluded environments, with mAP@50:95 stabilizing at 39%. Real-time testing confirmed the system's practical applicability and reliability under diverse environmental conditions. This solution represents a cost-effective, scalable approach to protecting rice fields, offering a significant leap forward in precision agriculture.