Application of the meanshift segmentation method for tree crown delineation from UAV imagery using Orfeo ToolBox

DOI: 10.35595/2414-9179-2025-2-31-154-163

View or download the article (Rus)

About the Authors

Leonid A. Shilonosov

Perm State University,
15, Bukireva str., Perm, 614068, Russia,
E-mail: lonya.sh@mail.ru

Eduard E. Rotermel

Perm State University,
15, Bukireva str., Perm, 614068, Russia,

Timur A. Belyaev

Perm State University,
15, Bukireva str., Perm, 614068, Russia,

Dmitriy V. Makurin

Perm State University,
15, Bukireva str., Perm, 614068, Russia,

Lyudmila A. Ivanchina

Perm State University,
15, Bukireva str., Perm, 614068, Russia,
E-mail: ivanchina.ludmila@yandex.ru

Abstract

Accurate automated detection of individual tree crowns from aerial imagery constitutes an important task. In recent years, fir stands in the Perm Krai have experienced mass dieback due to the impact of the Ussuri bark beetle (Polygraphus proximus Blandford). Effective management of such forests requires a system capable of recognizing tree crowns across different sanitary-condition categories: crowns with green needles, crowns with discolored needles, and trees lacking needles. The aim of this study was to evaluate the accuracy of recognizing individual crowns of fir stands affected by dieback from P. proximus using the MeanShift method. The object of investigation was a stand located within the Perm Forestry of Perm Krai. Since 2022, fir trees in this stand affected by P. proximus have been undergoing dieback. The proportion of fir in the stand composition is 30 %, and the proportion of standing dead fir killed by P. proximus is 71.6 %. Aerial surveys of the study area were conducted using a DJI Mavic 3M Multispectral UAV equipped with a multispectral camera and a DJI Mavic 3T (Thermal) UAV equipped with a thermal imaging camera. For crown delineation, the MeanShift segmentation tool was selected. Segmentation was performed in QGIS software using the Orfeo ToolBox module integrated into its analysis tools. The overall accuracy of crown recognition reached 73 %, and the F-measure value was 69 %. The method requires further refinement; a key limitation is its inability to segment every individual tree crown. The novelty of our crown-detection approach lies in the use of thermal imagery to distinguish crowns from the underlying surface.

Keywords

individual tree crowns, four-eyed fir bark beetle, orthophotomap, MeanShift Segmentation, Orfeo Tool Box module, thermal-imaging

References

  1. Anuchin N.P. Dendrometry. Moscow: Forest Industry, 1982. 552 p. (in Russian).
  2. Bunkova N.P., Zalesov S.V., Zalesova E.S., Magasumova A.G., Osipenko R.A. Basics of Phytomonitoring. Ekaterinburg: Ural State Forest Engineering University, 2020. 90 p. (in Russian).
  3. De Luca G., Silva J.M.N., Cerasoli S., Araújo J., Campos J., Di Fazio S., Modica G. Object-Based Land Cover Classification of Cork Oak Woodlands using UAV Imagery and Orfeo ToolBox. Remote Sensing, 2019. V. 11. P. 1238. DOI: 10.3390/rs11101238.
  4. Ecke S., Dempewolf J., Frey J., Schwaller A., Endres E., Klemmt H.-J., Tiede D., Seifert T. UAV-based Forest Health Monitoring: A Systematic Review. Remote Sensing, 2022. V. 14. No. 13. P. 3205. DOI: 10.3390/rs14133205.
  5. Fujimoto A., Haga C., Matsui T., Machimura T., Hayashi K., Sugita S., Takagi H. An End to End Process Development for UAV-SFM Based Forest Monitoring: Individual Tree Detection, Species Classification and Carbon Dynamics Simulation. Forests, 2019. V. 10. No. 8. P. 680. DOI: 10.3390/f10080680.
  6. Gommershtadt O.M. Using UAV Data to Study the Horizontal Structure of Northern Forests. Proceedings of the XXVI International Scientific Conference of Students, Postgraduates and Young Scientists “Lomonosov-2019”. Sevastopol: Branch of Lomonosov Moscow State University in the city of Sevastopol, 2019. P. 245–246 (in Russian).
  7. Kapil R., Marvasti-Zadeh S.M., Erbilgin N., Ray N. ShadowSense: Unsupervised Domain Adaptation and Feature Fusion for Shadow-Agnostic Tree Crown Detection from RGB-Thermal Drone Imagery. IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2024. DOI: 10.1109/WACV57701.2024.00808.
  8. Kapil R., Marvasti-Zadeh S.M., Goodsman D., Ray N., Erbilgin N. Classification of Bark Beetle-Induced Forest Tree Mortality Using Deep Learning. arXiv preprint arXiv:2207.07241, 2022. DOI: 10.48550/arXiv.2207.07241.
  9. Krivets S.A., Kerchev I.A., Bisirova E.M., Volkova E.S., Astapenko S.A., Efremenko A.A., Kosilov A.Yu., Kudryavtsev P.P., Kuznetzova Yu.P., Ponomarev V.I., Potapkin A.B., Taraskin E.G., Titova V.V., Shilonosov A.O., Baranchikov Yu.N. Overview of the Current Secondary Range of the Four-eyed Fir Bark Beetle (Polygraphus Proximus Blandford) in the Russian Federation. Russian Journal of Biological Invasions, 2024. No. 1. P. 49–69 (in Russian). DOI: 10.35885/1996-1499-17-1-49-69.
  10. Nikitina A.D. Automatic Segmentation of Tree Crowns in Pine Forests Using MASK R-CNN on RGB Imagery from UAVS. Forest Science Issues, 2024a. V. 7. No. 2. P. 1–21 (in Russian). DOI: 10.31509/2658-607x-202372-146.
  11. Nikitina A.D. Determination of Carbon Stocks in Pine Stands of Coniferous-Broadleaf Forests Using High-Resolution Survey Data. Dissertation for PhD in Biological Sciences. Moscow, 2024b. 204 p. (in Russian).
  12. Sukhikh V.I., Gusev N.N., Danyulis E.P. Aerial Methods in Forest Management. Moscow: Forestry Industry, 1977. 192 p. (in Russian).

For citation: Shilonosov L.A., Rotermel E.E., Belyaev T.A., Makurin D.V., Ivanchina L.A. Application of the meanshift segmentation method for tree crown delineation from UAV imagery using Orfeo ToolBox. InterCarto. InterGIS. Moscow: MSU, Faculty of Geography, 2025. V. 31. Part 2. P. 154–163. DOI: 10.35595/2414-9179-2025-2-31-154-163 (in Russian)