边缘增强的BECU-Net模型高分辨率遥感影像耕地提取

董张玉; 李金徽; 张晋; 于金秋; 安森

doi:10.11834/jrs.20222268

AI+遥感 | 浏览量 : 0 下载量: 304 CSCD: 0 更多指标

R-PDF
PDF
导出
分享
收藏
专辑

边缘增强的BECU-Net模型高分辨率遥感影像耕地提取
Cultivated land extraction from high-resolution remote sensing images based on BECU-Net model with edge enhancement
2023年27卷第12期页码：2847-2859
纸质出版日期： 2023-12-07 ，
DOI： 10.11834/jrs.20222268

扫描看全文

董张玉，李金徽，张晋，于金秋，安森.2023.边缘增强的BECU-Net模型高分辨率遥感影像耕地提取.遥感学报，27（12）： 2847-2859

Dong Z Y， Li J H， Zhang J， Yu J Q and An S. 2023. Cultivated land extraction from high-resolution remote sensing images based on BECU-Net model with edge enhancement. National Remote Sensing Bulletin， 27（12）：2847-2859
董张玉，李金徽，张晋，于金秋，安森.2023.边缘增强的BECU-Net模型高分辨率遥感影像耕地提取.遥感学报，27（12）： 2847-2859 DOI： 10.11834/jrs.20222268.

Dong Z Y， Li J H， Zhang J， Yu J Q and An S. 2023. Cultivated land extraction from high-resolution remote sensing images based on BECU-Net model with edge enhancement. National Remote Sensing Bulletin， 27（12）：2847-2859 DOI： 10.11834/jrs.20222268.

摘要

耕地作为国家粮食生产的重要保障，其空间分布是粮食安全评估、土地资源管理等领域的主要依据。为解决现有的耕地信息提取方法忽视地块的差异化特征和边缘细节蕴含的丰富信息，且提取结果碎片化、边界模糊问题。本研究以耕地为研究对象，采用一种结合EfficientNet骨干网络和U型框架构建的改进型耕地信息提取模型BECU-Net（Boundary Enhancement Classification U-Net），并为实现边缘特征和深度特征的信息互补，设计由CoT模块（Contextual Transformer Module）、门控卷积、scSE（Spatial-Channel Sequeeze and Excitation）注意力机制形成的边缘分支子网络，来提高模型处理边界信息的专注度。同时，构建含约束项的联合型边缘增强损失函数BE-LOSS（Boundary Enhancement Loss）进一步完善模型运算性能。使用GID高分二号RGB-NIR四波段数据，与梯度、指数、纹理特征图共同构建耕地特征机制。并分别与不同网络结构、不同损失函数的模型进行对比。结果表明：改进算法的总体精度和F1分数均有改善，相比于DeeplabV3+网络，提取精度提升2.24%，F1分数提升1.77%。本研究提出的新算法可为进一步解决耕地信息提取时边界模糊问题提供技术参考，为复杂交界的精准划分提供理论支撑。

Abstract

Cultivated land cover

as an important technical index to reflect the dynamic changes of human activities and the utilization degree of land resources

has been widely utilized in the fields of food security assessment and land management decision making. Existing information extraction methods ignore the differential characteristics of the plots and the rich information found in edge details

which results in fragmented extraction results with fuzzy boundaries. Therefore

an improved model that couples semantic segmentation model and edge enhancement is proposed to better solve the problem of insufficient fitting of cultivated land edges and fully utilize the rich semantic features and edge information in remote sensing images. The edge loss is designed accordingly to further improve the training accuracy and model performance.

We design an edge branching self-network formed by CoT unit

gated convolution

and SCSE attention mechanism to realize the information complementarity of edge and depth features. We construct a joint edge enhancement loss function called BE-loss with constraints to enhance the attention of the model to boundary information. On this basis

we construct a cultivated land information extraction model

that is

BECU-net

by combining the EfficientNet backbone network and U-frame. In the multi-feature input layer of this model

the index and texture features of the preprocessed data are pre-extracted

the input structure is adjusted

and the feature expression ability of the network is improved.

The extraction accuracy of cultivated land is 94.13%

and the F1-score is 95.17%. Compared with PANet

the extraction accuracy increased by 15.01%

and the F1-score improved by 7.93%. Compared with DeeplabV3+ network

the extraction accuracy is enhanced by 2.03%

and the F1-score is increased by 1.15%. The edge of cultivated land extracted by BECU-Net model is clear

and it is close to the real edge shape of cultivated land. Few holes and islands are observed. The extracted large parcels are not missing

and the edges and corners are sharp. The extracted small parcels have clear outlines and small deformation. At various gaps and complex edges

the extraction effect of GID dataset is significantly improved compared with that of the five other models. The effect is significant when used for edge extraction

The sawtooth and cavity phenomena of cultivated land patches are effectively restrained as well.

(1) The input layer of network structure with multiple features

including exponential features and texture features

can effectively reflect the characteristics of cultivated land. (2) The edge branch subnetwork focuses on processing the shape information to better identify the boundary details in the cultivated land image. Its edge features complement the depth features of the Efficient encoder

and they can be cascaded to fully utilize the shallow details. (3) The improved combined loss function called BE-Loss with regular term solves the problem of unbalanced training sample categories and non-edge pixel-dominated loss function. Overall

the algorithm in this study provides a technical reference for further solving the problem of fuzzy boundaries when extracting cultivated land information. It also offers theoretical support for the accurate division of complex boundaries.

关键词

遥感边缘增强耕地提取语义分割U-Net高分影像

Keywords

remote sensingedge enhancementcultivated land extractionsemantic segmentationU-Nethigh resolution remote sensing image

references

Bao Y T, Liu W, Gao O Y, Lin Z K and Hu Q. 2021. E-Unet++: a semantic segmentation method for remote sensing images//2021 IEEE 4th Advanced Information Management, Communicates, Electronic and Automation Control Conference. Chongqing: IEEE: 1858-1862 [DOI: 10.1109/IMCEC51613.2021.9482266http://dx.doi.org/10.1109/IMCEC51613.2021.9482266]

Chen L C, Zhu Y K, Papandreou G, Schroff F and Adam H. 2018. Encoder-decoder with atrous separable convolution for semantic image segmentation//Proceedings of the 15th European Conference on Computer vision. Munich: Springer: 833-851 [DOI: 10.1007/978-3-030-01234-2_49http://dx.doi.org/10.1007/978-3-030-01234-2_49]

Chen Z X, Wang L M, Wu W B, Jiang Z W and Li H. 2016. Monitoring plastic-mulched farmland by Landsat-8 OLI imagery using spectral and textural features. Remote Sensing, 8(4): 353 [DOI: 10.3390/rs8040353http://dx.doi.org/10.3390/rs8040353]

Du G M, Gai Z X and Wang H Y. 2021. Theoretical explanation and research framework of cultivated-land fragmentation in China. Journal of Earth Sciences and Environment, 43(6): 997-1008

杜国明, 盖兆雪, 王洪彦. 2021. 中国耕地细碎化的理论解析与研究框架. 地球科学与环境学报, 43(6): 997-1008 [DOI: 10.19814/j.jese.2021.07029http://dx.doi.org/10.19814/j.jese.2021.07029]

Fan D L, Su X Y, Weng B, Wang T S and Yang F Y. 2021. Research progress on remote sensing classification methods for farmland vegetation. AgriEngineering, 3(4): 971-989 [DOI: 10.3390/agriengineering3040061http://dx.doi.org/10.3390/agriengineering3040061]

Gan F P, Wang R S, Wang Y J and Fu Z W. 1999. The classification method based on remote sensing techniques for land use and cover. Remote Sensing for Land and Resources, 11(4): 40-45

甘甫平, 王润生, 王永江, 付正文. 1999. 基于遥感技术的土地利用与土地覆盖的分类方法. 国土资源遥感, 11(4): 40-45 [DOI: 10.6046/gtzyyg.1999.04.08http://dx.doi.org/10.6046/gtzyyg.1999.04.08]

Gong P and Howarth P J. 1990. The use of structural information for improving land-cover classification accuracies at the rural-urban fringe. Photogrammetric Engineering and Remote Sensing, 56(1): 67-73

He C, Li S L, Xiong D H, Fang P Z and Liao M S. 2020. Remote sensing image semantic segmentation based on edge information guidance. Remote Sensing, 12(9): 1501 [DOI: 10.3390/rs12091501http://dx.doi.org/10.3390/rs12091501]

Krizhevsky A, Sutskever I and Hinton G E. 2012. ImageNet classification with deep convolutional neural networks//Proceedings of the 25th International Conference on Neural Information Processing Systems. Lake Tahoe: Curran Associates Inc.: 1097-1105

Lee R Y, Chang K C, Ou D Y and Hsu C H. 2020. Evaluation of crop mapping on fragmented and complex slope farmlands through random forest and object-oriented analysis using unmanned aerial vehicles. Geocarto International, 35(12): 1293-1310 [DOI: 10.1080/10106049.2018.1559886http://dx.doi.org/10.1080/10106049.2018.1559886]

Li X Y, Sun X F, Meng Y X, Liang J J, Wu F and Li J W. 2020. Dice loss for data-imbalanced NLP tasks. arXiv preprint arXiv: 1911.02855 [DOI: 10.48550/arXiv.1911.02855http://dx.doi.org/10.48550/arXiv.1911.02855]

Li Y H, Yao T, Pan Y W and Mei T. 2023. Contextual transformer networks for visual recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(2): 1489-1500 [DOI: 10.1109/TPAMI.2022.3164083http://dx.doi.org/10.1109/TPAMI.2022.3164083]

Liu G Y, Song X and Lv J. 2015. Framland parcels extraction from high-resolution remote sensing images based on the two-stage image classification//Proceedings Volume 9812, MIPPR 2015: Automatic Target Recognition and Navigation. Enshi: SPIE: 239-244 [DOI: 10.1117/12.2209212http://dx.doi.org/10.1117/12.2209212]

Long J, Shelhamer E and Darrell T. 2015. Fully convolutional networks for semantic segmentation//2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE: 3431-3440 [DOI: 10.1109/CVPR.2015.7298965http://dx.doi.org/10.1109/CVPR.2015.7298965]

Ma M N. 1987. Basic theories and methods of remote sensing visual interpretation. Remote Sensing Information, (3): 26-29

马霭乃. 1987. 遥感目视解译的基本理论与方法. 遥感信息, (3): 26-29

Mather P M. 1986. Review of: “Introductory Digital Image Processing: A Remote Sensing Perspective” By J. R. JENSEN: (Englewood Cliffs, New Jersey: Prentice Hall, 1986) [Pp. 368.] Price £53-45. International Journal of Remote Sensing, 7(12): 1836-1838 [DOI: 10.1080/01431168608948975http://dx.doi.org/10.1080/01431168608948975]

Pan Y Z, Hu T G, Zhu X F, Zhang J S and Wang X D. 2012. Mapping cropland distributions using a hard and soft classification model. IEEE Transactions on Geoscience and Remote Sensing, 50(11): 4301-4312 [DOI: 10.1109/TGRS.2012.2193403http://dx.doi.org/10.1109/TGRS.2012.2193403]

Ronneberger O, Fischer P and Brox T. 2015. U-Net: convolutional networks for biomedical image segmentation//18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich: Springer: 234-241 [DOI: 10.1007/978-3-319-24574-4_28http://dx.doi.org/10.1007/978-3-319-24574-4_28]

Roy A G, Navab N and Wachinger C. 2018. Concurrent spatial and channel ‘squeeze and excitation’ in fully convolutional networks//21st International Conference on Medical Image Computing and Computer-Assisted Intervention. Granada: Springer: 421-429 [DOI: 10.1007/978-3-030-00928-1_48http://dx.doi.org/10.1007/978-3-030-00928-1_48]

Takikawa T, Acuna D, Jampani V and Fidler S. 2019. Gated-SCNN: gated shape CNNs for semantic segmentation. arXiv:1907.05740 [DOI: 10.48550/arXiv.1907.05740http://dx.doi.org/10.48550/arXiv.1907.05740]

Tan M X and Le Q V. 2020. EfficientNet: rethinking model scaling for convolutional neural networks. arXiv:1905.11946 [DOI: 10.48550/arXiv.1905.11946http://dx.doi.org/10.48550/arXiv.1905.11946]

Tong X Y, Xia G S, Lu Q K, Shen H F, Li S Y, You S C and Zhang L P. 2020. Land-cover classification with high-resolution remote sensing images using transferable deep models. Remote Sensing of Environment, 237: 111322 [DOI: 10.1016/j.rse.2019.111322http://dx.doi.org/10.1016/j.rse.2019.111322]

Verma A K, Garg P K and Prasad K S H. 2017. Sugarcane crop identification from LISS IV data using ISODATA, MLC, and indices based decision tree approach. Arabian Journal of Geosciences, 10(1): 16 [DOI: 10.1007/s12517-016-2815-xhttp://dx.doi.org/10.1007/s12517-016-2815-x]

Wu Y R, Wang W Z, Zhuang J X, Ma C F, Liu S H and Wu L Z. 2013. Extraction of saline land based on decision tree approach using Landsat TM DATA//2013 IEEE International Geoscience and Remote Sensing Symposium. Melbourne: IEEE: 3762-3765 [DOI: 10.1109/IGARSS.2013.6723649http://dx.doi.org/10.1109/IGARSS.2013.6723649]

Wu Z H, Lei S G, Bian Z F, Huang J and Zhang Y. 2019. Study of the desertification index based on the albedo-MSAVI feature space for semi-arid steppe region. Environmental Earth Sciences, 78(6): 232 [DOI: 10.1007/s12665-019-8111-9http://dx.doi.org/10.1007/s12665-019-8111-9]

Xia L G, Luo J C, Sun Y W and Yang H P. 2018. Deep extraction of cropland parcels from very high-resolution remotely sensed imagery//2018 7th International Conference on Agro-geoinformatics (Agro-geoinformatics). Hangzhou: IEEE: 1-5 [DOI: 10.1109/Agro-Geoinformatics.2018.8476002http://dx.doi.org/10.1109/Agro-Geoinformatics.2018.8476002]

Xu L, Ming D P, Zhou W, Bao H Q, Chen Y Y and Ling X. 2019. Farmland extraction from high spatial resolution remote sensing images based on stratified scale pre-estimation. Remote Sensing, 11(2): 108 [DOI: 10.3390/rs11020108http://dx.doi.org/10.3390/rs11020108]

Xu W N, Deng X P, Guo S X, Chen J S, Sun L Y, Zheng X R, Xiong Y F, Shen Y and Wang X Q. 2020. High-resolution u-net: preserving image details for cultivated land extraction. Sensors, 20(15): 4064 [DOI: 10.3390/s20154064http://dx.doi.org/10.3390/s20154064]

Yang B and Liu Y. 2008. Technological essentials of visual interpretation of agricultural remote sensing images. Modernizing Agriculture, (4): 37-39

杨博, 刘义. 2008. 农业遥感影像目视解译技术要点. 现代化农业, (4): 37-39 [DOI: 10.3969/j.issn.1001-0254.2008.04.003http://dx.doi.org/10.3969/j.issn.1001-0254.2008.04.003]

Yang X, Zhu D M, Yang R S, Fu Z T and Xie W B. 2020. A visible-band remote sensing index for extracting impervious surfaces. Transactions of the Chinese Society of Agricultural Engineering, 36(8): 127-134

杨栩, 朱大明, 杨润书, 付志涛, 谢文斌. 2020. 一种提取不透水面的可见光波段遥感指数. 农业工程学报, 36(8): 127-134 [DOI: 10.11975/j.issn.1002-6819.2020.08.016http://dx.doi.org/10.11975/j.issn.1002-6819.2020.08.016]

Zhang S L, Zhang H, Hu D B, Zeng D, Bian Z Y, Lu L J, Ma J H and Huang J. 2015. Edge-detecting operator-based selection of Huber regularization threshold for low-dose computed tomography imaging. Journal of Southern Medical University, 35(3): 375-379

张善立, 张华, 胡德斌, 曾栋, 边兆英, 路利军, 马建华, 黄静. 2015. 基于边缘检测算子的Huber正则化阈值选择方法在低剂量CT重建中的应用. 南方医科大学学报, 35(3): 375-379 [DOI: 10.3969/j.issn.1673-4254.2015.03.12http://dx.doi.org/10.3969/j.issn.1673-4254.2015.03.12]

Zheng C, Chen Y C, Shao J and Wang L G. 2022. An MRF-based multigranularity edge-preservation optimization for semantic segmentation of remote sensing images. IEEE Geoscience and Remote Sensing Letters, 19: 8008205 [DOI: 10.1109/LGRS.2021.3058939http://dx.doi.org/10.1109/LGRS.2021.3058939]

Zhou Z W, Siddiquee M M R, Tajbakhsh N and Liang J M. 2020. UNet++: redesigning skip connections to exploit multiscale features in image segmentation. IEEE Transactions on Medical Imaging, 39(6): 1856-1867 [DOI: 10.1109/TMI.2019.2959609http://dx.doi.org/10.1109/TMI.2019.2959609]

Zhu N Y, Liu C, Singer Z S, Danino T, Laine A F and Guo J. 2020. Segmentation with residual attention U-Net and an edge-enhancement approach preserves cell shape features. arXiv preprint arXiv: 2001.05548 [DOI: 10.48550/arXiv.2001.05548http://dx.doi.org/10.48550/arXiv.2001.05548]

文章被引用时，请邮件提醒。

提交

DeepLabv3+语义分割模型的济南市防尘绿网提取及时空变化分析

SARBuD 1.0：面向深度学习的GF-3 精细模式SAR建筑数据集

点云深度学习基准数据集

透视地球——新一代对地观测技术