UAV-Based, AI-powered detection and heat mapping of potential mosquito breeding sites using YOLOv8

Mosquito-borne diseases, particularly dengue, remain a major public health challenge in tropical regions such as the Philippines. In 2024, dengue cases increased sharply nationwide, with the Davao Region reporting over 9,000 cases and rising mortality. Conventional mosquito control methods are often labor-intensive, costly, and difficult to sustain, especially in underserved communities. This study proposes an automated, real-time mosquito breeding site detection system integrating unmanned aerial vehicles (UAVs) with artificial intelligence-based image analysis and spatial heat mapping. A DJI Air 3 drone conducts aerial surveillance, while an OKdo Jetson Nano microcontroller runs dual YOLOv8 models: YOLOv8n for object detection and YOLOv8seg for water segmentation, optimized using NVIDIA TensorRT. The system detects potential breeding containers such as buckets, tires, and stagnant water, and generates heat maps to classify risk levels across barangays in Caraga, Davao Oriental. Field tests over 72 flights showed inference times of 56.1-114.8 ms and frame rates of 13.21-15 FPS. The model achieved 93.20% accuracy with high recall, precision, and F1 scores. Regression analysis indicated a strong correlation (R2 = 0.937) between detected breeding sites and dengue cases, confirming predictive validity and supporting its use for targeted public health interventions.