Tree semantic segmentation from aerial image time series

Earth’s forests play an important role in the fight against climate change and are in turn negatively affected by it. Effective monitoring of different tree species is essential to understanding and improving the health and biodiversity of forests. In this work, we address the challenge of tree spec...

Full description

Saved in:
Bibliographic Details
Main Authors: Venkatesh Ramesh, Arthur Ouaknine, David Rolnick
Format: Article
Language:English
Published: Cambridge University Press 2025-01-01
Series:Environmental Data Science
Subjects:
Online Access:https://www.cambridge.org/core/product/identifier/S2634460225100137/type/journal_article
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849245897161965568
author Venkatesh Ramesh
Arthur Ouaknine
David Rolnick
author_facet Venkatesh Ramesh
Arthur Ouaknine
David Rolnick
author_sort Venkatesh Ramesh
collection DOAJ
description Earth’s forests play an important role in the fight against climate change and are in turn negatively affected by it. Effective monitoring of different tree species is essential to understanding and improving the health and biodiversity of forests. In this work, we address the challenge of tree species identification by performing tree crown semantic segmentation using an aerial image dataset spanning over a year. We compare models trained on single images versus those trained on time series to assess the impact of tree phenology on segmentation performance. We also introduce a simple convolutional block for extracting spatio-temporal features from image time series, enabling the use of popular pretrained backbones and methods. We leverage the hierarchical structure of tree species taxonomy by incorporating a custom loss function that refines predictions at three levels: species, genus, and higher-level taxa. Our best model achieves a mean Intersection over Union (mIoU) of 55.97%, outperforming single-image approaches particularly for deciduous trees where phenological changes are most noticeable. Our findings highlight the benefit of exploiting the time series modality via our Processor module. Furthermore, leveraging taxonomic information through our hierarchical loss function often, and in key cases significantly, improves semantic segmentation performance.
format Article
id doaj-art-e6b4241474e54cdf9903e89055977950
institution Kabale University
issn 2634-4602
language English
publishDate 2025-01-01
publisher Cambridge University Press
record_format Article
series Environmental Data Science
spelling doaj-art-e6b4241474e54cdf9903e890559779502025-08-20T03:58:40ZengCambridge University PressEnvironmental Data Science2634-46022025-01-01410.1017/eds.2025.10013Tree semantic segmentation from aerial image time seriesVenkatesh Ramesh0https://orcid.org/0009-0005-5400-2516Arthur Ouaknine1https://orcid.org/0000-0003-1090-6204David Rolnick2Mila, https://ror.org/05c22rx21Quebec AI Institute, Montréal, QC, Canada Département d’informatique et de recherche opérationnelle, https://ror.org/0161xgx34Université de Montréal, Montréal, QC, CanadaMila, https://ror.org/05c22rx21Quebec AI Institute, Montréal, QC, Canada School of Computer Science, https://ror.org/01pxwe438 McGill University , Montréal, QC, CanadaMila, https://ror.org/05c22rx21Quebec AI Institute, Montréal, QC, Canada School of Computer Science, https://ror.org/01pxwe438 McGill University , Montréal, QC, CanadaEarth’s forests play an important role in the fight against climate change and are in turn negatively affected by it. Effective monitoring of different tree species is essential to understanding and improving the health and biodiversity of forests. In this work, we address the challenge of tree species identification by performing tree crown semantic segmentation using an aerial image dataset spanning over a year. We compare models trained on single images versus those trained on time series to assess the impact of tree phenology on segmentation performance. We also introduce a simple convolutional block for extracting spatio-temporal features from image time series, enabling the use of popular pretrained backbones and methods. We leverage the hierarchical structure of tree species taxonomy by incorporating a custom loss function that refines predictions at three levels: species, genus, and higher-level taxa. Our best model achieves a mean Intersection over Union (mIoU) of 55.97%, outperforming single-image approaches particularly for deciduous trees where phenological changes are most noticeable. Our findings highlight the benefit of exploiting the time series modality via our Processor module. Furthermore, leveraging taxonomic information through our hierarchical loss function often, and in key cases significantly, improves semantic segmentation performance.https://www.cambridge.org/core/product/identifier/S2634460225100137/type/journal_articledeep learningforest monitoringphenologyremote sensingtime series
spellingShingle Venkatesh Ramesh
Arthur Ouaknine
David Rolnick
Tree semantic segmentation from aerial image time series
Environmental Data Science
deep learning
forest monitoring
phenology
remote sensing
time series
title Tree semantic segmentation from aerial image time series
title_full Tree semantic segmentation from aerial image time series
title_fullStr Tree semantic segmentation from aerial image time series
title_full_unstemmed Tree semantic segmentation from aerial image time series
title_short Tree semantic segmentation from aerial image time series
title_sort tree semantic segmentation from aerial image time series
topic deep learning
forest monitoring
phenology
remote sensing
time series
url https://www.cambridge.org/core/product/identifier/S2634460225100137/type/journal_article
work_keys_str_mv AT venkateshramesh treesemanticsegmentationfromaerialimagetimeseries
AT arthurouaknine treesemanticsegmentationfromaerialimagetimeseries
AT davidrolnick treesemanticsegmentationfromaerialimagetimeseries