Altitude-Optimized UAV surveillance for mesquite detection using YOLOv8

Invasive Prosopis juliflora (mesquite) trees have proliferated across Oman’s Dhofar region, threatening native ecosystems and water resources. Traditional ground surveys and satellite monitoring of this invasive species are labor-intensive and often lack sufficient resolution to identify individual trees in Dhofar’s expansive, rugged terrain. This study proposes an integrated Unmanned Aerial Vehicle (UAV) and deep learning approach to efficiently detect and map mesquite infestations, with a focus on evaluating the impact of UAV flight altitude on detection performance. Using a DJI Mavic 2 Pro drone, high-resolution aerial images were collected from multiple altitudes (10 m, 20 m, 50 m) over affected rangeland plots. Thousands of image patches were manually labeled to train a YOLOv8 object detection model. The trained model achieved high precision in identifying mesquite trees, with the lowest flight altitude yielding the most accurate detections due to greater image detail. Moderate altitudes provided a balance between coverage area and detection accuracy, informing an optimal surveillance strategy. This research highlights the novelty of comparing multi-altitude UAV imagery for ecological monitoring, demonstrating that integrating UAVs and artificial intelligence can significantly enhance invasive species management. The findings offer a scalable and impactful framework for protecting vulnerable ecosystems in Dhofar and similar environments.