Remote sensing classification of wetlands in regions around the South China Sea based on bilinear graph convolutional network

doi:10.13474/j.cnki.11-2246.2023.0134

Abstract

Abstract: Wetlands have important carbon sink functions, and play a key role in purifying water quality and regulating climate. There exist abundant wetland resources in regions around the South China Sea, and it is of great significance to monitor wetlands in this area to promote the cross-border joint protection of coastal wetlands in China and achieve the goal of carbon dioxide peaking and carbon neutrality. This paper proposes a wetland classification method based on BiGCN using object-oriented hierarchical classification. Random forest is used to distinguish wetlands from non-wetlands firstly, and then wetlands are sub-classified by the BiGCN. In the BiGCN, the methods of constructing bilinear model, optimizing graph structure and using better activation function are used to further optimize network performance. The results show that the overall classification accuracy of the proposed model is above 92% on the three Sentinel-2 data sets around the South China Sea, which is more than 4% higher than that of the existing graph convolution network, and the time consumption is greatly reduced.

Key words: wetland, classification, remote sensing, deep learning, South China Sea

CLC Number:

P237

LI Xinyuan, HE Zhi, LOU Anjun, XIAO Man. Remote sensing classification of wetlands in regions around the South China Sea based on bilinear graph convolutional network[J]. Bulletin of Surveying and Mapping, 2023, 0(5): 44-50.

References

[1] MA K, LIU J. Changes in soil organic carbon stocks of wetlands on China's Zoige Plateau from 1980 to 2010[J]. Ecological Modelling, 2016, 327:18-28.
[2] 王勇辉. "21世纪海上丝绸之路"东南亚战略支点国家的构建[J]. 世界经济与政治论坛, 2016(3):61-73.
[3] 赵海迪, 刘世梁, 董世魁, 等. 三江源区人类干扰与湿地空间变化关系研究[J]. 湿地科学, 2014, 12(1):22-28.
[4] HE Zhi, LIU Han, WANG Yiwen, et al. Generative adversarial networks-based semi-supervised learning for hyperspectral image classification[J]. Remote Sensing, 2017, 9(10):1042.
[5] 贾文翰, 刘越岩. 基于卷积神经网络的高分遥感影像内陆盐沼湿地信息提取[J]. 测绘通报, 2021(6):89-92.
[6] JAMALI A, MAHDIANPARI M, MOHAMMADIMANESH F, et al. A synergic use of sentinel-1 and sentinel-2 imagery for complex wetland classification using generative adversarial network (GAN) scheme[J]. Water, 2021, 13(24):3601.
[7] 程晓晖, 李长辉, 欧佳斌, 等. 深度学习在海珠国家湿地公园统一确权调查中的应用[J]. 测绘通报, 2021(11):110-114.
[8] KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL].[2016-01-01]. https://arxiv.org/abs/1609.02907.
[9] LIN T Y, ROYCHOWDHURY A, MAJI S. Bilinear CNN models for fine-grained visual recognition[C]//Proceedings of 2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile:IEEE, 2016:1449-1457.
[10] 袁杰. 从海岸带到岛屿带:"环中国海"的地理与生态[J]. 南方文物, 2015(2):87-93.
[11] 刘云刚, 刘玄宇, 张争胜. 渔民视角下中国南海的领域构建[J]. 地理科学, 2020, 40(7):1062-1071.
[12] GONG Peng, LIU Han, ZHANG Meinan, et al. Stable classification with limited sample:transferring a 30-m resolution sample set collected in 2015 to mapping 10-m resolution global land cover in 2017[J]. Science Bulletin, 2019, 64(6):370-373.
[13] BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1):5-32.
[14] HAMMOND D K, VANDERGHEYNST P, GRIBONVAL R. Wavelets on graphs via spectral graph theory[J]. Applied and Computational Harmonic Analysis, 2011, 30(2):129-150.
[15] CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3):273-297.
[16] 李朝奎, 方文, 董小姣. 面向对象和规则的高分辨率影像分类研究[J]. 测绘通报, 2015(9):9-13.