Global, multi-scale standing deadwood segmentation in centimeter-scale aerial images

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F67985939%3A_____%2F25%3A00641584" target="_blank" >RIV/67985939:_____/25:00641584 - isvavai.cz</a>

Result on the web

<a href="https://doi.org/10.1016/j.ophoto.2025.100104" target="_blank" >https://doi.org/10.1016/j.ophoto.2025.100104</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.ophoto.2025.100104" target="_blank" >10.1016/j.ophoto.2025.100104</a>

Result language

angličtina

Original language name

Global, multi-scale standing deadwood segmentation in centimeter-scale aerial images

Original language description

With tree mortality rates rising across many regions of the world, efficient methods to map dead trees are becoming increasingly important to monitor forest dieback, assess ecological impacts, and guide management strategies. Deep learning-based pattern recognition combined with the high spatial detail of aerial images from drones or airplanes provides an avenue for mapping dead tree crowns or partial canopy dieback, collectively referred to as standing deadwood. However, current methods for mapping standing deadwood are limited to specific biomes or image resolutions. Here, we present a transformer-based semantic segmentation model that generalizes across forest biomes and a wide range of image resolutions (1-28 cm) for mapping both dead tree crowns and partial canopy dieback. Our approach combines a SegFormer-based transformer architecture for image feature extraction and Focal Tversky Loss to mitigate class imbalance. We used a globally distributed crowd-sourced dataset of 434 high-resolution aerial images and manual delineations of standing deadwood of vastly varying quality. The orthophotos span all major forest biomes and cover 10,778 hectares. To further mitigate imbalances across biomes, resolutions, deadwood occurrence, and image sources, we developed a fourdimensional sampling scheme that ensures balanced representation during training. The models were trained and evaluated using heterogeneous crowd-sourced data, which, as expected, negatively affects the F1-scores. A visual inspection on independent data highlights the very precise quality of the segmentation. Our analysis revealed resolution-dependent performance variations across biomes, suggesting a relationship between optimal mapping resolution and biome-specific characteristics. We make both our model and a machine-learning-ready dataset publicly available on deadtrees.earth to support future research in tree mortality mapping.

Czech name

—

Czech description

—

Type

J_imp - Article in a specialist periodical, which is included in the Web of Science database

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

<a href="/en/project/GA23-05272S" target="_blank" >GA23-05272S: Tropical cyclone activity, drivers, and impact on forest ecosystems at different spatial and temporal scales</a><br>

Continuities

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year

2025

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Name of the periodical

ISPRS Open Journal of Photogrammetry and Remote Sensing

ISSN

2667-3932

e-ISSN

2667-3932

Volume of the periodical

18

Issue of the periodical within the volume

DEC 2025

Country of publishing house

NL - THE KINGDOM OF THE NETHERLANDS

Number of pages

12

Pages from-to

100104

UT code for WoS article

001608953400001

EID of the result in the Scopus database

2-s2.0-105020985955