Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks
Accurate extraction of farmland boundaries is crucial for improving the efficiency of farmland surveys, achieving precise agricultural management, enhancing farmers’ production conditions, protecting the ecological environment, and promoting local economic development. Remote sensing and deep learni...
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-10-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/15/20/4900 |
| _version_ | 1797572400390340608 |
|---|---|
| author | Chen Long Song Wenlong Sun Tao Lu Yizhu Jiang Wei Liu Jun Liu Hongjie Feng Tianshi Gui Rongjie Haider Abbas Meng Lingwei Lin Shengjie He Qian |
| author_facet | Chen Long Song Wenlong Sun Tao Lu Yizhu Jiang Wei Liu Jun Liu Hongjie Feng Tianshi Gui Rongjie Haider Abbas Meng Lingwei Lin Shengjie He Qian |
| author_sort | Chen Long |
| collection | DOAJ |
| description | Accurate extraction of farmland boundaries is crucial for improving the efficiency of farmland surveys, achieving precise agricultural management, enhancing farmers’ production conditions, protecting the ecological environment, and promoting local economic development. Remote sensing and deep learning are feasible methods for creating large-scale farmland boundary maps. However, existing neural network models have limitations that restrict the accuracy and reliability of agricultural parcel extraction using remote sensing technology. In this study, we used high-resolution satellite images (2 m, 1 m, and 0.8 m) and the U<sup>2</sup>-Net++ model based on the RSU module, deep separable convolution, and the channel-spatial attention mechanism module to extract different types of fields. Our model exhibited significant improvements in farmland parcel extraction compared with the other models. It achieved an F1-score of 97.13%, which is a 7.36% to 17.63% improvement over older models such as U-Net and FCN and a more than 2% improvement over advanced models such as DeepLabv3+ and U<sup>2</sup>-Net. These results indicate that U<sup>2</sup>-Net++ holds the potential for widespread application in the production of large-scale farmland boundary maps. |
| first_indexed | 2024-03-10T20:55:43Z |
| format | Article |
| id | doaj.art-7b0b73aa94704213af10b9ae91ae00cd |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T20:55:43Z |
| publishDate | 2023-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-7b0b73aa94704213af10b9ae91ae00cd2023-11-19T17:58:00ZengMDPI AGRemote Sensing2072-42922023-10-011520490010.3390/rs15204900Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural NetworksChen Long0Song Wenlong1Sun Tao2Lu Yizhu3Jiang Wei4Liu Jun5Liu Hongjie6Feng Tianshi7Gui Rongjie8Haider Abbas9Meng Lingwei10Lin Shengjie11He Qian12State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaSuqian City Sucheng District Water Conservancy Bureau, Suqian 223800, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaCollege of Resource Environment and Tourism, Capital Normal University, Beijing 100048, ChinaState Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, ChinaCollege of Resource Environment and Tourism, Capital Normal University, Beijing 100048, ChinaAccurate extraction of farmland boundaries is crucial for improving the efficiency of farmland surveys, achieving precise agricultural management, enhancing farmers’ production conditions, protecting the ecological environment, and promoting local economic development. Remote sensing and deep learning are feasible methods for creating large-scale farmland boundary maps. However, existing neural network models have limitations that restrict the accuracy and reliability of agricultural parcel extraction using remote sensing technology. In this study, we used high-resolution satellite images (2 m, 1 m, and 0.8 m) and the U<sup>2</sup>-Net++ model based on the RSU module, deep separable convolution, and the channel-spatial attention mechanism module to extract different types of fields. Our model exhibited significant improvements in farmland parcel extraction compared with the other models. It achieved an F1-score of 97.13%, which is a 7.36% to 17.63% improvement over older models such as U-Net and FCN and a more than 2% improvement over advanced models such as DeepLabv3+ and U<sup>2</sup>-Net. These results indicate that U<sup>2</sup>-Net++ holds the potential for widespread application in the production of large-scale farmland boundary maps.https://www.mdpi.com/2072-4292/15/20/4900high-resolution remote sensing imagerydepth-wise separable convolutionchannel-spatial attention mechanismdeep learningfarmland parcel extraction |
| spellingShingle | Chen Long Song Wenlong Sun Tao Lu Yizhu Jiang Wei Liu Jun Liu Hongjie Feng Tianshi Gui Rongjie Haider Abbas Meng Lingwei Lin Shengjie He Qian Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks Remote Sensing high-resolution remote sensing imagery depth-wise separable convolution channel-spatial attention mechanism deep learning farmland parcel extraction |
| title | Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks |
| title_full | Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks |
| title_fullStr | Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks |
| title_full_unstemmed | Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks |
| title_short | Field Patch Extraction Based on High-Resolution Imaging and U<sup>2</sup>-Net++ Convolutional Neural Networks |
| title_sort | field patch extraction based on high resolution imaging and u sup 2 sup net convolutional neural networks |
| topic | high-resolution remote sensing imagery depth-wise separable convolution channel-spatial attention mechanism deep learning farmland parcel extraction |
| url | https://www.mdpi.com/2072-4292/15/20/4900 |
| work_keys_str_mv | AT chenlong fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT songwenlong fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT suntao fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT luyizhu fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT jiangwei fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT liujun fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT liuhongjie fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT fengtianshi fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT guirongjie fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT haiderabbas fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT menglingwei fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT linshengjie fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks AT heqian fieldpatchextractionbasedonhighresolutionimagingandusup2supnetconvolutionalneuralnetworks |