Technical Handbook: HD Drone Tool for AI-Based Tree Biostructural Detection and Forest Inventory

This technical handbook presents the DigiMedFor HD Drone tool, an integrated drone-based forest inventory and monitoring system developed for Digital Mediterranean Forestry applications. The document describes the full workflow for detecting individual tree crowns from UAV imagery, estimating key biostructural parameters such as crown area, tree height, diameter at breast height, and stem volume, and georeferencing each detection through image metadata and WGS-84 coordinate conversion. It also explains the use of multi-model YOLO architectures, multispectral image fusion, allometric estimation pipelines, disease pre-classification, PostgreSQL database integration, and interactive geospatial analysis for forest management and environmental monitoring.

The handbook is intended for forest rangers, environmental scientists, research institutions, conservation organizations, and other stakeholders interested in operational AI tools for forestry. In addition to the technical methodology, it documents the Kavala pilot experience, system validation, database structure, visualization tools, and the broader environmental, social, ethical, and policy dimensions of deploying AI-assisted decision-support systems in Mediterranean forest ecosystems. The publication serves as a comprehensive reference for understanding, applying, and further developing the DigiMedFor drone-based forest inventory framework.

Also, the metrics of the used YOLO pipeline in detecting tree crowns is included in this post's repository along with the metrics of the used YOLO models' inference in ground images.

A selected set of HD drone images taken from the Limnia's Kavala polygon in a zip format is included. Finally, as we also tested synthetic panchromatic images, the created repository of images based on drone's HD images and their corresponding multispectral images is also uploaded as a zip file in this post.