A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information
Accurately extracting pixel-level buildings from high-resolution remote sensing images is significant for various geographical information applications. Influenced by different natural, cultural, and social development levels, buildings may vary in shape and distribution, making it difficult for the...
| Principais autores: | , , , , , |
|---|---|
| Formato: | Artigo |
| Idioma: | English |
| Publicado em: |
MDPI AG
2024-02-01
|
| coleção: | Sensors |
| Assuntos: | |
| Acesso em linha: | https://www.mdpi.com/1424-8220/24/3/1006 |
| _version_ | 1827354517732065280 |
|---|---|
| author | Zhaojun Pang Rongming Hu Wu Zhu Renyi Zhu Yuxin Liao Xiying Han |
| author_facet | Zhaojun Pang Rongming Hu Wu Zhu Renyi Zhu Yuxin Liao Xiying Han |
| author_sort | Zhaojun Pang |
| collection | DOAJ |
| description | Accurately extracting pixel-level buildings from high-resolution remote sensing images is significant for various geographical information applications. Influenced by different natural, cultural, and social development levels, buildings may vary in shape and distribution, making it difficult for the network to maintain a stable segmentation effect of buildings in different areas of the image. In addition, the complex spectra of features in remote sensing images can affect the extracted details of multi-scale buildings in different ways. To this end, this study selects parts of Xi’an City, Shaanxi Province, China, as the study area. A parallel encoded building extraction network (MARS-Net) incorporating multiple attention mechanisms is proposed. MARS-Net builds its parallel encoder through DCNN and transformer to take advantage of their extraction of local and global features. According to the different depth positions of the network, coordinate attention (CA) and convolutional block attention module (CBAM) are introduced to bridge the encoder and decoder to retain richer spatial and semantic information during the encoding process, and adding the dense atrous spatial pyramid pooling (DenseASPP) captures multi-scale contextual information during the upsampling of the layers of the decoder. In addition, a spectral information enhancement module (SIEM) is designed in this study. SIEM further enhances building segmentation by blending and enhancing multi-band building information with relationships between bands. The experimental results show that MARS-Net performs better extraction results and obtains more effective enhancement after adding SIEM. The IoU on the self-built Xi’an and WHU building datasets are 87.53% and 89.62%, respectively, while the respective F1 scores are 93.34% and 94.52%. |
| first_indexed | 2024-03-08T03:48:45Z |
| format | Article |
| id | doaj.art-5fd95c26032a45fba00e74951b3794ed |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-08T03:48:45Z |
| publishDate | 2024-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-5fd95c26032a45fba00e74951b3794ed2024-02-09T15:22:33ZengMDPI AGSensors1424-82202024-02-01243100610.3390/s24031006A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral InformationZhaojun Pang0Rongming Hu1Wu Zhu2Renyi Zhu3Yuxin Liao4Xiying Han5School of Geomatics, Xi’an University of Science and Technology, Xi’an 710054, ChinaSchool of Geomatics, Xi’an University of Science and Technology, Xi’an 710054, ChinaSchool of Geological Engineering and Geomatics, Chang’an University, Xi’an 710054, ChinaThe First Institute of Geoinformation Mapping, Ministry of Natural Resources, Xi’an 710054, ChinaSchool of Geomatics, Xi’an University of Science and Technology, Xi’an 710054, ChinaThe First Institute of Geoinformation Mapping, Ministry of Natural Resources, Xi’an 710054, ChinaAccurately extracting pixel-level buildings from high-resolution remote sensing images is significant for various geographical information applications. Influenced by different natural, cultural, and social development levels, buildings may vary in shape and distribution, making it difficult for the network to maintain a stable segmentation effect of buildings in different areas of the image. In addition, the complex spectra of features in remote sensing images can affect the extracted details of multi-scale buildings in different ways. To this end, this study selects parts of Xi’an City, Shaanxi Province, China, as the study area. A parallel encoded building extraction network (MARS-Net) incorporating multiple attention mechanisms is proposed. MARS-Net builds its parallel encoder through DCNN and transformer to take advantage of their extraction of local and global features. According to the different depth positions of the network, coordinate attention (CA) and convolutional block attention module (CBAM) are introduced to bridge the encoder and decoder to retain richer spatial and semantic information during the encoding process, and adding the dense atrous spatial pyramid pooling (DenseASPP) captures multi-scale contextual information during the upsampling of the layers of the decoder. In addition, a spectral information enhancement module (SIEM) is designed in this study. SIEM further enhances building segmentation by blending and enhancing multi-band building information with relationships between bands. The experimental results show that MARS-Net performs better extraction results and obtains more effective enhancement after adding SIEM. The IoU on the self-built Xi’an and WHU building datasets are 87.53% and 89.62%, respectively, while the respective F1 scores are 93.34% and 94.52%.https://www.mdpi.com/1424-8220/24/3/1006high-resolution remote sensing imagerybuilding extractiondeep convolutional neural network (DCNN)transformerspectral enhancement |
| spellingShingle | Zhaojun Pang Rongming Hu Wu Zhu Renyi Zhu Yuxin Liao Xiying Han A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information Sensors high-resolution remote sensing imagery building extraction deep convolutional neural network (DCNN) transformer spectral enhancement |
| title | A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information |
| title_full | A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information |
| title_fullStr | A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information |
| title_full_unstemmed | A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information |
| title_short | A Building Extraction Method for High-Resolution Remote Sensing Images with Multiple Attentions and Parallel Encoders Combining Enhanced Spectral Information |
| title_sort | building extraction method for high resolution remote sensing images with multiple attentions and parallel encoders combining enhanced spectral information |
| topic | high-resolution remote sensing imagery building extraction deep convolutional neural network (DCNN) transformer spectral enhancement |
| url | https://www.mdpi.com/1424-8220/24/3/1006 |
| work_keys_str_mv | AT zhaojunpang abuildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT rongminghu abuildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT wuzhu abuildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT renyizhu abuildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT yuxinliao abuildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT xiyinghan abuildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT zhaojunpang buildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT rongminghu buildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT wuzhu buildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT renyizhu buildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT yuxinliao buildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation AT xiyinghan buildingextractionmethodforhighresolutionremotesensingimageswithmultipleattentionsandparallelencoderscombiningenhancedspectralinformation |