Deep Semantic Segmentation of Trees Using Multispectral Images
Forests can be efficiently monitored by automatic semantic segmentation of trees using satellite and/or aerial images. Still, several challenges can make the problem difficult, including the varying spectral signature of different trees, lack of sufficient labelled data, and geometrical occlusions....
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9872072/ |
| _version_ | 1828147874178793472 |
|---|---|
| author | Irem Ulku Erdem Akagunduz Pedram Ghamisi |
| author_facet | Irem Ulku Erdem Akagunduz Pedram Ghamisi |
| author_sort | Irem Ulku |
| collection | DOAJ |
| description | Forests can be efficiently monitored by automatic semantic segmentation of trees using satellite and/or aerial images. Still, several challenges can make the problem difficult, including the varying spectral signature of different trees, lack of sufficient labelled data, and geometrical occlusions. In this article, we address the tree segmentation problem using multispectral imagery. While we carry out large-scale experiments on several deep learning architectures using various spectral input combinations, we also attempt to explore whether hand-crafted spectral vegetation indices can improve the performance of deep learning models in the segmentation of trees. Our experiments include benchmarking a variety of multispectral remote sensing image sets, deep semantic segmentation architectures, and various spectral bands as inputs, including a number of hand-crafted spectral vegetation indices. From our large-scale experiments, we draw several useful conclusions. One particularly important conclusion is that, with no additional computation burden, combining different categories of multispectral vegetation indices, such as NVDI, atmospherically resistant vegetation index, and soil-adjusted vegetation index, within a single three-channel input, and using the state-of-the-art semantic segmentation architectures, tree segmentation accuracy can be improved under certain conditions, compared to using high-resolution visible and/or near-infrared input. |
| first_indexed | 2024-04-11T21:06:43Z |
| format | Article |
| id | doaj.art-32290d4c2d5548aa958fd8537763cc65 |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-04-11T21:06:43Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-32290d4c2d5548aa958fd8537763cc652022-12-22T04:03:15ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352022-01-01157589760410.1109/JSTARS.2022.32031459872072Deep Semantic Segmentation of Trees Using Multispectral ImagesIrem Ulku0https://orcid.org/0000-0003-4998-607XErdem Akagunduz1https://orcid.org/0000-0002-0792-7306Pedram Ghamisi2https://orcid.org/0000-0003-1203-741XDepartment of Computer Engineering, Ankara University, Ankara, TurkeyGraduate School of Informatics, Middle East Technical University, Ankara, TurkeyInstitute of Advanced Research in Artificial Intelligence, Vienna, AustriaForests can be efficiently monitored by automatic semantic segmentation of trees using satellite and/or aerial images. Still, several challenges can make the problem difficult, including the varying spectral signature of different trees, lack of sufficient labelled data, and geometrical occlusions. In this article, we address the tree segmentation problem using multispectral imagery. While we carry out large-scale experiments on several deep learning architectures using various spectral input combinations, we also attempt to explore whether hand-crafted spectral vegetation indices can improve the performance of deep learning models in the segmentation of trees. Our experiments include benchmarking a variety of multispectral remote sensing image sets, deep semantic segmentation architectures, and various spectral bands as inputs, including a number of hand-crafted spectral vegetation indices. From our large-scale experiments, we draw several useful conclusions. One particularly important conclusion is that, with no additional computation burden, combining different categories of multispectral vegetation indices, such as NVDI, atmospherically resistant vegetation index, and soil-adjusted vegetation index, within a single three-channel input, and using the state-of-the-art semantic segmentation architectures, tree segmentation accuracy can be improved under certain conditions, compared to using high-resolution visible and/or near-infrared input.https://ieeexplore.ieee.org/document/9872072/Satellite imagerysemantic segmentationtree segmentationvegetation indices (VIs) |
| spellingShingle | Irem Ulku Erdem Akagunduz Pedram Ghamisi Deep Semantic Segmentation of Trees Using Multispectral Images IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Satellite imagery semantic segmentation tree segmentation vegetation indices (VIs) |
| title | Deep Semantic Segmentation of Trees Using Multispectral Images |
| title_full | Deep Semantic Segmentation of Trees Using Multispectral Images |
| title_fullStr | Deep Semantic Segmentation of Trees Using Multispectral Images |
| title_full_unstemmed | Deep Semantic Segmentation of Trees Using Multispectral Images |
| title_short | Deep Semantic Segmentation of Trees Using Multispectral Images |
| title_sort | deep semantic segmentation of trees using multispectral images |
| topic | Satellite imagery semantic segmentation tree segmentation vegetation indices (VIs) |
| url | https://ieeexplore.ieee.org/document/9872072/ |
| work_keys_str_mv | AT iremulku deepsemanticsegmentationoftreesusingmultispectralimages AT erdemakagunduz deepsemanticsegmentationoftreesusingmultispectralimages AT pedramghamisi deepsemanticsegmentationoftreesusingmultispectralimages |