Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images
Due to the inconsistent spatiotemporal spectral scales, a remote sensing dataset over a large-scale area and over long-term time series will have large variations and large statistical distribution features, which will lead to a performance drop of the deep learning model that is only trained on the...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-03-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/5/1235 |
| _version_ | 1797473836291063808 |
|---|---|
| author | Zhenchao Tang Calvin Yu-Chian Chen Chengzhen Jiang Dongying Zhang Weiran Luo Zhiming Hong Huaiwei Sun |
| author_facet | Zhenchao Tang Calvin Yu-Chian Chen Chengzhen Jiang Dongying Zhang Weiran Luo Zhiming Hong Huaiwei Sun |
| author_sort | Zhenchao Tang |
| collection | DOAJ |
| description | Due to the inconsistent spatiotemporal spectral scales, a remote sensing dataset over a large-scale area and over long-term time series will have large variations and large statistical distribution features, which will lead to a performance drop of the deep learning model that is only trained on the source domain. For building an extraction task, deep learning methods perform weak generalization from the source domain to the other domain. To solve the problem, we propose a Capsule–Encoder–Decoder model. We use a vector named capsule to store the characteristics of the building and its parts. In our work, the encoder extracts capsules from remote sensing images. Capsules contain the information of the buildings’ parts. Additionally, the decoder calculates the relationship between the target building and its parts. The decoder corrects the buildings’ distribution and up-samples them to extract target buildings. Using remote sensing images in the lower Yellow River as the source dataset, building extraction experiments were trained on both our method and the mainstream methods. Compared with the mainstream methods on the source dataset, our method achieves convergence faster, and our method shows higher accuracy. Significantly, without fine tuning, our method can reduce the error rates of building extraction results on an almost unfamiliar dataset. The building parts’ distribution in capsules has high-level semantic information, and capsules can describe the characteristics of buildings more comprehensively, which are more explanatory. The results prove that our method can not only effectively extract buildings but also perform great generalization from the source remote sensing dataset to another. |
| first_indexed | 2024-03-09T20:23:02Z |
| format | Article |
| id | doaj.art-2709cd07133b4ca08de05042d3b3ba5b |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T20:23:02Z |
| publishDate | 2022-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-2709cd07133b4ca08de05042d3b3ba5b2023-11-23T23:43:31ZengMDPI AGRemote Sensing2072-42922022-03-01145123510.3390/rs14051235Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing ImagesZhenchao Tang0Calvin Yu-Chian Chen1Chengzhen Jiang2Dongying Zhang3Weiran Luo4Zhiming Hong5Huaiwei Sun6School of Civil and Hydraulic Engineering, Huazhong University of Science and Technology, Wuhan 430074, ChinaSchool of Intelligent Systems Engineering, Sun Yat-sen University, Shenzhen 518107, ChinaCenter of Geographic Information, Yellow River Engineering Consulting Co., Ltd., Zhengzhou 450003, ChinaSchool of Civil and Hydraulic Engineering, Huazhong University of Science and Technology, Wuhan 430074, ChinaSchool of Water Science and Engineering, Zhengzhou University, Zhengzhou 450001, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, ChinaSchool of Civil and Hydraulic Engineering, Huazhong University of Science and Technology, Wuhan 430074, ChinaDue to the inconsistent spatiotemporal spectral scales, a remote sensing dataset over a large-scale area and over long-term time series will have large variations and large statistical distribution features, which will lead to a performance drop of the deep learning model that is only trained on the source domain. For building an extraction task, deep learning methods perform weak generalization from the source domain to the other domain. To solve the problem, we propose a Capsule–Encoder–Decoder model. We use a vector named capsule to store the characteristics of the building and its parts. In our work, the encoder extracts capsules from remote sensing images. Capsules contain the information of the buildings’ parts. Additionally, the decoder calculates the relationship between the target building and its parts. The decoder corrects the buildings’ distribution and up-samples them to extract target buildings. Using remote sensing images in the lower Yellow River as the source dataset, building extraction experiments were trained on both our method and the mainstream methods. Compared with the mainstream methods on the source dataset, our method achieves convergence faster, and our method shows higher accuracy. Significantly, without fine tuning, our method can reduce the error rates of building extraction results on an almost unfamiliar dataset. The building parts’ distribution in capsules has high-level semantic information, and capsules can describe the characteristics of buildings more comprehensively, which are more explanatory. The results prove that our method can not only effectively extract buildings but also perform great generalization from the source remote sensing dataset to another.https://www.mdpi.com/2072-4292/14/5/1235remote sensing imagesbuilding extractioncapsule–encoder–decoderexplainabilitygeneralization |
| spellingShingle | Zhenchao Tang Calvin Yu-Chian Chen Chengzhen Jiang Dongying Zhang Weiran Luo Zhiming Hong Huaiwei Sun Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images Remote Sensing remote sensing images building extraction capsule–encoder–decoder explainability generalization |
| title | Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images |
| title_full | Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images |
| title_fullStr | Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images |
| title_full_unstemmed | Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images |
| title_short | Capsule–Encoder–Decoder: A Method for Generalizable Building Extraction from Remote Sensing Images |
| title_sort | capsule encoder decoder a method for generalizable building extraction from remote sensing images |
| topic | remote sensing images building extraction capsule–encoder–decoder explainability generalization |
| url | https://www.mdpi.com/2072-4292/14/5/1235 |
| work_keys_str_mv | AT zhenchaotang capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages AT calvinyuchianchen capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages AT chengzhenjiang capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages AT dongyingzhang capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages AT weiranluo capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages AT zhiminghong capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages AT huaiweisun capsuleencoderdecoderamethodforgeneralizablebuildingextractionfromremotesensingimages |