A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation
Applying deep-learning methods, especially fully convolutional networks (FCNs), has become a popular option for land-cover classification or segmentation in remote sensing. Compared with traditional solutions, these approaches have shown promising generalization capabilities and precision levels in...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2019-05-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/11/9/1051 |
| _version_ | 1831692539594276864 |
|---|---|
| author | Guangming Wu Yimin Guo Xiaoya Song Zhiling Guo Haoran Zhang Xiaodan Shi Ryosuke Shibasaki Xiaowei Shao |
| author_facet | Guangming Wu Yimin Guo Xiaoya Song Zhiling Guo Haoran Zhang Xiaodan Shi Ryosuke Shibasaki Xiaowei Shao |
| author_sort | Guangming Wu |
| collection | DOAJ |
| description | Applying deep-learning methods, especially fully convolutional networks (FCNs), has become a popular option for land-cover classification or segmentation in remote sensing. Compared with traditional solutions, these approaches have shown promising generalization capabilities and precision levels in various datasets of different scales, resolutions, and imaging conditions. To achieve superior performance, a lot of research has focused on constructing more complex or deeper networks. However, using an ensemble of different fully convolutional models to achieve better generalization and to prevent overfitting has long been ignored. In this research, we design four stacked fully convolutional networks (SFCNs), and a feature alignment framework for multi-label land-cover segmentation. The proposed feature alignment framework introduces an alignment loss of features extracted from basic models to balance their similarity and variety. Experiments on a very high resolution(VHR) image dataset with six categories of land-covers indicates that the proposed SFCNs can gain better performance when compared to existing deep learning methods. In the 2nd variant of SFCN, the optimal feature alignment gains increments of 4.2% (0.772 vs. 0.741), 6.8% (0.629 vs. 0.589), and 5.5% (0.727 vs. 0.689) for its f1-score, jaccard index, and kappa coefficient, respectively. |
| first_indexed | 2024-12-20T11:55:45Z |
| format | Article |
| id | doaj.art-e1d1bdd801ad4667b7247f6283534599 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-12-20T11:55:45Z |
| publishDate | 2019-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-e1d1bdd801ad4667b7247f62835345992022-12-21T19:41:40ZengMDPI AGRemote Sensing2072-42922019-05-01119105110.3390/rs11091051rs11091051A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover SegmentationGuangming Wu0Yimin Guo1Xiaoya Song2Zhiling Guo3Haoran Zhang4Xiaodan Shi5Ryosuke Shibasaki6Xiaowei Shao7Center for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanCenter for Spatial Information Science, University of Tokyo, Kashiwa 277-8568, JapanApplying deep-learning methods, especially fully convolutional networks (FCNs), has become a popular option for land-cover classification or segmentation in remote sensing. Compared with traditional solutions, these approaches have shown promising generalization capabilities and precision levels in various datasets of different scales, resolutions, and imaging conditions. To achieve superior performance, a lot of research has focused on constructing more complex or deeper networks. However, using an ensemble of different fully convolutional models to achieve better generalization and to prevent overfitting has long been ignored. In this research, we design four stacked fully convolutional networks (SFCNs), and a feature alignment framework for multi-label land-cover segmentation. The proposed feature alignment framework introduces an alignment loss of features extracted from basic models to balance their similarity and variety. Experiments on a very high resolution(VHR) image dataset with six categories of land-covers indicates that the proposed SFCNs can gain better performance when compared to existing deep learning methods. In the 2nd variant of SFCN, the optimal feature alignment gains increments of 4.2% (0.772 vs. 0.741), 6.8% (0.629 vs. 0.589), and 5.5% (0.727 vs. 0.689) for its f1-score, jaccard index, and kappa coefficient, respectively.https://www.mdpi.com/2072-4292/11/9/1051land-cover classificationimage segmentationensemble learingfeature alignmentfully convolutional networks |
| spellingShingle | Guangming Wu Yimin Guo Xiaoya Song Zhiling Guo Haoran Zhang Xiaodan Shi Ryosuke Shibasaki Xiaowei Shao A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation Remote Sensing land-cover classification image segmentation ensemble learing feature alignment fully convolutional networks |
| title | A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation |
| title_full | A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation |
| title_fullStr | A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation |
| title_full_unstemmed | A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation |
| title_short | A Stacked Fully Convolutional Networks with Feature Alignment Framework for Multi-Label Land-cover Segmentation |
| title_sort | stacked fully convolutional networks with feature alignment framework for multi label land cover segmentation |
| topic | land-cover classification image segmentation ensemble learing feature alignment fully convolutional networks |
| url | https://www.mdpi.com/2072-4292/11/9/1051 |
| work_keys_str_mv | AT guangmingwu astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT yiminguo astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT xiaoyasong astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT zhilingguo astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT haoranzhang astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT xiaodanshi astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT ryosukeshibasaki astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT xiaoweishao astackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT guangmingwu stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT yiminguo stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT xiaoyasong stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT zhilingguo stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT haoranzhang stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT xiaodanshi stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT ryosukeshibasaki stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation AT xiaoweishao stackedfullyconvolutionalnetworkswithfeaturealignmentframeworkformultilabellandcoversegmentation |