AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network
Building extraction from high-resolution remote sensing images has various applications, such as urban planning and population estimation. However, buildings have intraclass heterogeneity and interclass homogeneity in high-resolution remote sensing images with complex backgrounds, which makes the ac...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-07-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/23/14/6349 |
| _version_ | 1827731854870970368 |
|---|---|
| author | Weizhi Liu Haixin Liu Chao Liu Junjie Kong Can Zhang |
| author_facet | Weizhi Liu Haixin Liu Chao Liu Junjie Kong Can Zhang |
| author_sort | Weizhi Liu |
| collection | DOAJ |
| description | Building extraction from high-resolution remote sensing images has various applications, such as urban planning and population estimation. However, buildings have intraclass heterogeneity and interclass homogeneity in high-resolution remote sensing images with complex backgrounds, which makes the accurate extraction of building instances challenging and regular building boundaries difficult to maintain. In this paper, an attention-gated and direction-field-optimized building instance extraction network (AGDF-Net) is proposed. Two refinements are presented, including an Attention-Gated Feature Pyramid Network (AG-FPN) and a Direction Field Optimization Module (DFOM), which are used to improve information flow and optimize the mask, respectively. The AG-FPN promotes complementary semantic and detail information by measuring information importance to control the addition of low-level and high-level features. The DFOM predicts the pixel-level direction field of each instance and iteratively corrects the direction field based on the initial segmentation. Experimental results show that the proposed method outperforms the six state-of-the-art instance segmentation methods and three semantic segmentation methods. Specifically, AGDF-Net improves the objective-level metric AP and the pixel-level metric IoU by 1.1%~9.4% and 3.55%~5.06%. |
| first_indexed | 2024-03-11T00:40:37Z |
| format | Article |
| id | doaj.art-fc7ee99857ea445ca067859cf11b0f32 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-11T00:40:37Z |
| publishDate | 2023-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-fc7ee99857ea445ca067859cf11b0f322023-11-18T21:16:29ZengMDPI AGSensors1424-82202023-07-012314634910.3390/s23146349AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction NetworkWeizhi Liu0Haixin Liu1Chao Liu2Junjie Kong3Can Zhang4College of Mining and Geomatics, Hebei University of Engineering, Handan 056038, ChinaCollege of Mining and Geomatics, Hebei University of Engineering, Handan 056038, ChinaCollege of Mining and Geomatics, Hebei University of Engineering, Handan 056038, ChinaCollege of Mining and Geomatics, Hebei University of Engineering, Handan 056038, ChinaCollege of Mining and Geomatics, Hebei University of Engineering, Handan 056038, ChinaBuilding extraction from high-resolution remote sensing images has various applications, such as urban planning and population estimation. However, buildings have intraclass heterogeneity and interclass homogeneity in high-resolution remote sensing images with complex backgrounds, which makes the accurate extraction of building instances challenging and regular building boundaries difficult to maintain. In this paper, an attention-gated and direction-field-optimized building instance extraction network (AGDF-Net) is proposed. Two refinements are presented, including an Attention-Gated Feature Pyramid Network (AG-FPN) and a Direction Field Optimization Module (DFOM), which are used to improve information flow and optimize the mask, respectively. The AG-FPN promotes complementary semantic and detail information by measuring information importance to control the addition of low-level and high-level features. The DFOM predicts the pixel-level direction field of each instance and iteratively corrects the direction field based on the initial segmentation. Experimental results show that the proposed method outperforms the six state-of-the-art instance segmentation methods and three semantic segmentation methods. Specifically, AGDF-Net improves the objective-level metric AP and the pixel-level metric IoU by 1.1%~9.4% and 3.55%~5.06%.https://www.mdpi.com/1424-8220/23/14/6349building extractioninstance segmentationattention gatedirection field |
| spellingShingle | Weizhi Liu Haixin Liu Chao Liu Junjie Kong Can Zhang AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network Sensors building extraction instance segmentation attention gate direction field |
| title | AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network |
| title_full | AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network |
| title_fullStr | AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network |
| title_full_unstemmed | AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network |
| title_short | AGDF-Net: Attention-Gated and Direction-Field-Optimized Building Instance Extraction Network |
| title_sort | agdf net attention gated and direction field optimized building instance extraction network |
| topic | building extraction instance segmentation attention gate direction field |
| url | https://www.mdpi.com/1424-8220/23/14/6349 |
| work_keys_str_mv | AT weizhiliu agdfnetattentiongatedanddirectionfieldoptimizedbuildinginstanceextractionnetwork AT haixinliu agdfnetattentiongatedanddirectionfieldoptimizedbuildinginstanceextractionnetwork AT chaoliu agdfnetattentiongatedanddirectionfieldoptimizedbuildinginstanceextractionnetwork AT junjiekong agdfnetattentiongatedanddirectionfieldoptimizedbuildinginstanceextractionnetwork AT canzhang agdfnetattentiongatedanddirectionfieldoptimizedbuildinginstanceextractionnetwork |