The article presents a study on the development of a comprehensive methodology for automatic tree detection based on the YOLO neural network architecture. The paper describes in detail the full data processing cycle, from the preparation of a labeled dataset with bounding boxes to the training and validation of the model. Special attention is given to the creation of a training dataset and the analysis of model performance metrics (mAP, Precision, Recall) on a test set.
It is shown that using predicted bounding boxes as masks for object detection in point clouds allows for the visualization of individual trees. The results confirm the potential of the proposed methodology for automated forest monitoring and inventory tasks, providing a foundation for further development of digitalization methods in forestry.

Christina A. Martin
Аssistant
St. Petersburg State Forestry University
5, Institutskii per., St. Petersburg, 194021, Russia
Lecturer
Russian State Hydrometeorological University
79, Voronezhskaya Str., St. Petersburg, 192007, Russia
е-mail: martynchris@mail.ru
SPIN-code: 7779-6962
AuthorID: 1187677

Irma A. Martin
Candidate of Technical Sciences
Associate Professor of the Department of Applied Informatics
Russian State Hydrometeorological University
79, Voronezhskaya Str., St. Petersburg, 192007, Russia
е-mail: irma_martyn@mail.ru
ORCID: 0000-0002-4332-7308
SPIN-code: 9386-0716
AuthorID: 1004119

Evgeny P. Istomin
Doctor of Technical Sciences, Professor
Director of the Institute of Information Systems and Geotechnologies
Russian State Hydrometeorological University
79, Voronezhskaya Str., St. Petersburg, 192007, Russia
e-mail: biom220@bk.ru
ORCID: 0000-0001-6247-4373
Scopus ID: 56951051300
SPIN-code: 6404-9070
AuthorID: 333123

Yaroslav A. Petrov
Candidate of Technical Sciences, Associate Professor
Associate Professor of the Department of Applied Informatics
Russian State Hydrometeorological University
79, Voronezhskaya Str., Saint-Petersburg, 192007, Russia
е-mail: ORCID: 0000-0002-9185-441X
SPIN-code: 4170-3003
AuthorID: 899415

1. Martyn I.A., Martyn K.A. Automation of the process of accounting forest area using machine learning and AI methods. In: Informatsionnye sistemy i tekhnologii : sbornik nauchnykh trudov. Iss. 17. Saint-Petersburg : SPBGLTU im. S.M. Kirova; 2025. pp. 131–133.
2. Weinstein B.G., Marconi S., Bohlman S., Zare A., White E. Individual Tree-Crown Detection in RGB Imagery Using Semi-Supervised Deep Learning Neural Networks. Remote Sensing. 2019;11(11):1309. DOI: 10.3390/rs11111309.
3. Vagizov M.R., Vitlev K.A., Popov A.I., Makarov K.A., Samsonov V.A., Sokolov A.G. Imagery for managing geoinformation models of forest ecosystems. In: Modelirovanie i situatsionnoe upravlenie kachestvom slozhnykh sistem: sbornik dokladov Chetvertoi Vserossiiskoi nauchnoi konferentsii (Saint-Petersburg, 18–22 April 2023). Saint-Petersburg : GUAP, 2023. pp. 131–134.
4. Ultralytics YOLO documentation. Available at: https://docs.ultralytics.com/ (accessed 20.01.2026).
5. Liu G.R. Machine learning with Python. Theory and Applications. Singapore : World Scientific; 2023. 693 p.
6. Gallatin K., Albon C. Machine Learning with Python Cookbook. Practical solutions from preprocessing to deep learning. Sebastopol : O’Reilly, 2023. 366 p.
7. Solawetz J., Francesco. What is YOLOv8? A Complete Guid. Roboflow. 23.10.2024. Available at: https://blog.roboflow.com/what-is-yolov8/ (accessed 20.01.2026).