Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset
<p>Image-based grain sizing has been used to measure grain size more efficiently compared with traditional methods (e.g., sieving and Wolman pebble count). However, current methods to automatically detect individual grains are largely based on detecting grain interstices from image intensity w...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2022-04-01
|
| Series: | Earth Surface Dynamics |
| Online Access: | https://esurf.copernicus.org/articles/10/349/2022/esurf-10-349-2022.pdf |
| _version_ | 1828366297914671104 |
|---|---|
| author | X. Chen X. Chen M. A. Hassan X. Fu |
| author_facet | X. Chen X. Chen M. A. Hassan X. Fu |
| author_sort | X. Chen |
| collection | DOAJ |
| description | <p>Image-based grain sizing has been used to measure grain size more
efficiently compared with traditional methods (e.g., sieving and Wolman pebble
count). However, current methods to automatically detect individual grains
are largely based on detecting grain interstices from image intensity which
not only require a significant level of expertise for parameter tuning but
also underperform when they are applied to suboptimal environments (e.g.,
dense organic debris, various sediment lithology). We proposed a model
(GrainID) based on convolutional neural networks to measure grain size in a
diverse range of fluvial environments. A dataset of more than 125 000
grains from flume and field measurements were compiled to develop GrainID.
Tests were performed to compare the predictive ability of GrainID with
sieving, manual labeling, Wolman pebble counts (Wolman, 1954) and
BASEGRAIN (Detert and Weitbrecht, 2012). When compared with the sieving
results for a sandy-gravel bed, GrainID yielded high predictive accuracy
(comparable to the performance of manual labeling) and outperformed
BASEGRAIN and Wolman pebble counts (especially for small grains). For the
entire evaluation dataset, GrainID once again showed fewer predictive errors
and significantly lower variation in results in comparison with BASEGRAIN and
Wolman pebble counts and maintained this advantage even in uncalibrated
rivers with drone images. Moreover, the existence of vegetation and noise
have little influence on the performance of GrainID. Analysis indicated that
GrainID performed optimally when the image resolution is higher than 1.8 mm pixel<span class="inline-formula"><sup>−1</sup></span>, the image tile size is <span class="inline-formula">512×512</span> pixels and the grain area
truncation values (the area of smallest detectable grains) were equal to 18–25 pixels.</p> |
| first_indexed | 2024-04-14T05:43:20Z |
| format | Article |
| id | doaj.art-c2c95619fb8c4d88b85cdea5bdcfd97d |
| institution | Directory Open Access Journal |
| issn | 2196-6311 2196-632X |
| language | English |
| last_indexed | 2024-04-14T05:43:20Z |
| publishDate | 2022-04-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Earth Surface Dynamics |
| spelling | doaj.art-c2c95619fb8c4d88b85cdea5bdcfd97d2022-12-22T02:09:22ZengCopernicus PublicationsEarth Surface Dynamics2196-63112196-632X2022-04-011034936610.5194/esurf-10-349-2022Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne datasetX. Chen0X. Chen1M. A. Hassan2X. Fu3Department of Hydraulic Engineering, State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing, ChinaDepartment of Geography, The University of British Columbia, Vancouver, BC, CanadaDepartment of Geography, The University of British Columbia, Vancouver, BC, CanadaDepartment of Hydraulic Engineering, State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing, China<p>Image-based grain sizing has been used to measure grain size more efficiently compared with traditional methods (e.g., sieving and Wolman pebble count). However, current methods to automatically detect individual grains are largely based on detecting grain interstices from image intensity which not only require a significant level of expertise for parameter tuning but also underperform when they are applied to suboptimal environments (e.g., dense organic debris, various sediment lithology). We proposed a model (GrainID) based on convolutional neural networks to measure grain size in a diverse range of fluvial environments. A dataset of more than 125 000 grains from flume and field measurements were compiled to develop GrainID. Tests were performed to compare the predictive ability of GrainID with sieving, manual labeling, Wolman pebble counts (Wolman, 1954) and BASEGRAIN (Detert and Weitbrecht, 2012). When compared with the sieving results for a sandy-gravel bed, GrainID yielded high predictive accuracy (comparable to the performance of manual labeling) and outperformed BASEGRAIN and Wolman pebble counts (especially for small grains). For the entire evaluation dataset, GrainID once again showed fewer predictive errors and significantly lower variation in results in comparison with BASEGRAIN and Wolman pebble counts and maintained this advantage even in uncalibrated rivers with drone images. Moreover, the existence of vegetation and noise have little influence on the performance of GrainID. Analysis indicated that GrainID performed optimally when the image resolution is higher than 1.8 mm pixel<span class="inline-formula"><sup>−1</sup></span>, the image tile size is <span class="inline-formula">512×512</span> pixels and the grain area truncation values (the area of smallest detectable grains) were equal to 18–25 pixels.</p>https://esurf.copernicus.org/articles/10/349/2022/esurf-10-349-2022.pdf |
| spellingShingle | X. Chen X. Chen M. A. Hassan X. Fu Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset Earth Surface Dynamics |
| title | Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset |
| title_full | Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset |
| title_fullStr | Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset |
| title_full_unstemmed | Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset |
| title_short | Convolutional neural networks for image-based sediment detection applied to a large terrestrial and airborne dataset |
| title_sort | convolutional neural networks for image based sediment detection applied to a large terrestrial and airborne dataset |
| url | https://esurf.copernicus.org/articles/10/349/2022/esurf-10-349-2022.pdf |
| work_keys_str_mv | AT xchen convolutionalneuralnetworksforimagebasedsedimentdetectionappliedtoalargeterrestrialandairbornedataset AT xchen convolutionalneuralnetworksforimagebasedsedimentdetectionappliedtoalargeterrestrialandairbornedataset AT mahassan convolutionalneuralnetworksforimagebasedsedimentdetectionappliedtoalargeterrestrialandairbornedataset AT xfu convolutionalneuralnetworksforimagebasedsedimentdetectionappliedtoalargeterrestrialandairbornedataset |