Automated Wildlife Recognition Using OpenCV for Conservation at Rinjani National Park

This study presents the design and validation of an AI-based camera-trap system for wildlife monitoring at the border of Rinjani National Park, Indonesia. The system uses the YOLOv4 framework integrated with OpenCV to detect long-tailed macaques (Macaca fascicularis) in real-time. A total of 269 annotated images were used, including 202 for training, 17 for validation, and 50 for testing. The model was trained using Google Colaboratory and achieved a detection accuracy of 92.83%. Image pre-processing and labeling were conducted via Roboflow, and the model was optimized for potential deployment on a Raspberry Pi platform. Although physical deployment was not conducted, the system design supports low-power embedded implementation for field use. The results indicate that the proposed method can reliably detect camouflaged and partially occluded monkeys, suggesting its potential for mitigating human–wildlife conflict through smart conservation technology.