Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery
In recent history, normalized digital surface models (nDSMs) have been constantly gaining importance as a means to solve large-scale geographic problems. High-resolution surface models are precious, as they can provide detailed information for a specific area. However, measurements with a high resol...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2023-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/10189905/ |
_version_ | 1797330394595459072 |
---|---|
author | Konstantin Muller Robert Leppich Christian Geis Vanessa Borst Patrick Aravena Pelizari Samuel Kounev Hannes Taubenbock |
author_facet | Konstantin Muller Robert Leppich Christian Geis Vanessa Borst Patrick Aravena Pelizari Samuel Kounev Hannes Taubenbock |
author_sort | Konstantin Muller |
collection | DOAJ |
description | In recent history, normalized digital surface models (nDSMs) have been constantly gaining importance as a means to solve large-scale geographic problems. High-resolution surface models are precious, as they can provide detailed information for a specific area. However, measurements with a high resolution are time consuming and costly. Only a few approaches exist to create high-resolution nDSMs for extensive areas. This article explores approaches to extract high-resolution nDSMs from low-resolution Sentinel-2 data, allowing us to derive large-scale models. We thereby utilize the advantages of Sentinel 2 being open access, having global coverage, and providing steady updates through a high repetition rate. Several deep learning models are trained to overcome the gap in producing high-resolution surface maps from low-resolution input data. With U-Net as a base architecture, we extend the capabilities of our model by integrating tailored multiscale encoders with differently sized kernels in the convolution as well as conformed self-attention inside the skip connection gates. Using pixelwise regression, our U-Net base models can achieve a mean height error of approximately 2 m. Moreover, through our enhancements to the model architecture, we reduce the model error by more than 7%. |
first_indexed | 2024-03-08T07:19:14Z |
format | Article |
id | doaj.art-42d2531ae8c74178a693e0c7d075bd6c |
institution | Directory Open Access Journal |
issn | 2151-1535 |
language | English |
last_indexed | 2024-03-08T07:19:14Z |
publishDate | 2023-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
spelling | doaj.art-42d2531ae8c74178a693e0c7d075bd6c2024-02-03T00:01:30ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352023-01-01168508851910.1109/JSTARS.2023.329771010189905Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 ImageryKonstantin Muller0https://orcid.org/0000-0001-6540-3124Robert Leppich1https://orcid.org/0000-0003-4711-7743Christian Geis2https://orcid.org/0000-0002-7961-8553Vanessa Borst3https://orcid.org/0009-0004-7123-7934Patrick Aravena Pelizari4https://orcid.org/0000-0003-0984-4675Samuel Kounev5https://orcid.org/0000-0001-9742-2063Hannes Taubenbock6https://orcid.org/0000-0003-4360-9126Department of Computer Science, Julius-Maximilians-Universität Würzburg, Würzburg, GermanyDepartment of Computer Science, Julius-Maximilians-Universität Würzburg, Würzburg, GermanyGerman Remote Sensing Data Center, German Aerospace Center, Weßling, GermanyDepartment of Computer Science, Julius-Maximilians-Universität Würzburg, Würzburg, GermanyGerman Remote Sensing Data Center, German Aerospace Center, Weßling, GermanyDepartment of Computer Science, Julius-Maximilians-Universität Würzburg, Würzburg, GermanyGerman Remote Sensing Data Center, German Aerospace Center, Weßling, GermanyIn recent history, normalized digital surface models (nDSMs) have been constantly gaining importance as a means to solve large-scale geographic problems. High-resolution surface models are precious, as they can provide detailed information for a specific area. However, measurements with a high resolution are time consuming and costly. Only a few approaches exist to create high-resolution nDSMs for extensive areas. This article explores approaches to extract high-resolution nDSMs from low-resolution Sentinel-2 data, allowing us to derive large-scale models. We thereby utilize the advantages of Sentinel 2 being open access, having global coverage, and providing steady updates through a high repetition rate. Several deep learning models are trained to overcome the gap in producing high-resolution surface maps from low-resolution input data. With U-Net as a base architecture, we extend the capabilities of our model by integrating tailored multiscale encoders with differently sized kernels in the convolution as well as conformed self-attention inside the skip connection gates. Using pixelwise regression, our U-Net base models can achieve a mean height error of approximately 2 m. Moreover, through our enhancements to the model architecture, we reduce the model error by more than 7%.https://ieeexplore.ieee.org/document/10189905/Deep learningmultiscale encodersentinelsurface model |
spellingShingle | Konstantin Muller Robert Leppich Christian Geis Vanessa Borst Patrick Aravena Pelizari Samuel Kounev Hannes Taubenbock Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Deep learning multiscale encoder sentinel surface model |
title | Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery |
title_full | Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery |
title_fullStr | Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery |
title_full_unstemmed | Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery |
title_short | Deep Neural Network Regression for Normalized Digital Surface Model Generation With Sentinel-2 Imagery |
title_sort | deep neural network regression for normalized digital surface model generation with sentinel 2 imagery |
topic | Deep learning multiscale encoder sentinel surface model |
url | https://ieeexplore.ieee.org/document/10189905/ |
work_keys_str_mv | AT konstantinmuller deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery AT robertleppich deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery AT christiangeis deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery AT vanessaborst deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery AT patrickaravenapelizari deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery AT samuelkounev deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery AT hannestaubenbock deepneuralnetworkregressionfornormalizeddigitalsurfacemodelgenerationwithsentinel2imagery |