GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information
Road extraction from high-resolution remote-sensing images has high application values in various fields. However, such work is susceptible to the influence of the surrounding environment due to the diverse slenderness and complex connectivity of roads, leading to false judgment and omission during...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/21/5476 |
| _version_ | 1827645574610944000 |
|---|---|
| author | Zixuan Zhang Xuan Sun Yuxi Liu |
| author_facet | Zixuan Zhang Xuan Sun Yuxi Liu |
| author_sort | Zixuan Zhang |
| collection | DOAJ |
| description | Road extraction from high-resolution remote-sensing images has high application values in various fields. However, such work is susceptible to the influence of the surrounding environment due to the diverse slenderness and complex connectivity of roads, leading to false judgment and omission during extraction. To solve this problem, a road-extraction network, the global attention multi-path dilated convolution gated refinement Network (GMR-Net), is proposed. The GMR-Net is facilitated by both local and global information. A residual module with an attention mechanism is first designed to obtain global and other aggregate information for each location’s features. Then, a multi-path dilated convolution (MDC) approach is used to extract road features at different scales, i.e., to achieve multi-scale road feature extraction. Finally, gated refinement units (GR) are proposed to filter out ambiguous features for the gradual refinement of details. Multiple road-extraction methods are compared in this study using the Deep-Globe and Massachusetts datasets. Experiments on these two datasets demonstrate that the proposed method achieves F1-scores of 87.38 and 85.70%, respectively, outperforming other approaches on segmentation accuracy and generalization ability. |
| first_indexed | 2024-03-09T18:42:04Z |
| format | Article |
| id | doaj.art-6f70dbb787324ba49a69131e8fb0a5c3 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T18:42:04Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-6f70dbb787324ba49a69131e8fb0a5c32023-11-24T06:39:44ZengMDPI AGRemote Sensing2072-42922022-10-011421547610.3390/rs14215476GMR-Net: Road-Extraction Network Based on Fusion of Local and Global InformationZixuan Zhang0Xuan Sun1Yuxi Liu2Zhou Enlai School of Government, Nankai University, Tianjin 300350, ChinaZhou Enlai School of Government, Nankai University, Tianjin 300350, ChinaCollege of Electronic Information and Optical Engineering, Nankai University, Tianjin 300350, ChinaRoad extraction from high-resolution remote-sensing images has high application values in various fields. However, such work is susceptible to the influence of the surrounding environment due to the diverse slenderness and complex connectivity of roads, leading to false judgment and omission during extraction. To solve this problem, a road-extraction network, the global attention multi-path dilated convolution gated refinement Network (GMR-Net), is proposed. The GMR-Net is facilitated by both local and global information. A residual module with an attention mechanism is first designed to obtain global and other aggregate information for each location’s features. Then, a multi-path dilated convolution (MDC) approach is used to extract road features at different scales, i.e., to achieve multi-scale road feature extraction. Finally, gated refinement units (GR) are proposed to filter out ambiguous features for the gradual refinement of details. Multiple road-extraction methods are compared in this study using the Deep-Globe and Massachusetts datasets. Experiments on these two datasets demonstrate that the proposed method achieves F1-scores of 87.38 and 85.70%, respectively, outperforming other approaches on segmentation accuracy and generalization ability.https://www.mdpi.com/2072-4292/14/21/5476road extractionattention mechanismmulti-scale featurelocal and global information |
| spellingShingle | Zixuan Zhang Xuan Sun Yuxi Liu GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information Remote Sensing road extraction attention mechanism multi-scale feature local and global information |
| title | GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information |
| title_full | GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information |
| title_fullStr | GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information |
| title_full_unstemmed | GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information |
| title_short | GMR-Net: Road-Extraction Network Based on Fusion of Local and Global Information |
| title_sort | gmr net road extraction network based on fusion of local and global information |
| topic | road extraction attention mechanism multi-scale feature local and global information |
| url | https://www.mdpi.com/2072-4292/14/21/5476 |
| work_keys_str_mv | AT zixuanzhang gmrnetroadextractionnetworkbasedonfusionoflocalandglobalinformation AT xuansun gmrnetroadextractionnetworkbasedonfusionoflocalandglobalinformation AT yuxiliu gmrnetroadextractionnetworkbasedonfusionoflocalandglobalinformation |