As part of the RAIDO Horizon Europe initiative, researchers from MetaMind Innovations – MINDS, Kingston University, International Hellenic University, and Aristotle University of Thessaloniki (AUTH) have developed an innovative tree detection methodology based on machine learning algorithms, specifically designed to identify individual trees in orchard environments from UAV imagery.
This research advances European agricultural technology by utilizing unmanned aerial vehicles equipped with multispectral cameras to efficiently capture orthomosaic images of extensive cultivation areas within minutes. The research team conducted a comprehensive evaluation of two state-of-the-art object detection algorithms—Detectron2 and YOLOv8—for precise tree identification and mask generation, achieving an exceptional F1-Score of 94.85% in cherry tree detection scenarios.
To further enhance detection precision, the team implemented an OTSU thresholding method which significantly improved mask accuracy. This approach achieved 85.30% on Intersection over Union (IoU) metrics, substantially outperforming both Detectron2 (79.83%) and YOLOv8 (75.36%) in crown extraction accuracy.
This advancement supports the European Union’s sustainable agriculture goals by enabling more precise resource management and reduced environmental impact through targeted interventions.
Research Team: Vasileios Moysiadis, Ilias Siniosoglou, George Kokkonis, Vasilis A., Thomas Lagkas, Sotirios Goudos, Panagiotis Sarigiannidis
Published on the prestigious MDPI Agriculture 2024 journal
Read the full paper: https://www.mdpi.com/2077-0472/14/2/322
