图Transformer支持下的河网模式识别

doi:10.11947/j. AGCS.2024.20230582.

摘要/Abstract

摘要：

河网模式识别在地形地貌分析、地质矿物探测、河网数据多尺度变换等研究中发挥重要作用。为克服基于形态特征与几何特征的空间统计方法的稳健性不足问题，引进图卷积神经网络是当前的主要手段，然而图卷积方法仅关注河网形态的局部特征，仍未实现从全局视角出发的河网模式识别决策。因此，本文提出了一种图Transformer支持下的河网模式识别方法。该方法在河网几何形态知识支持下利用对偶图思想构建河网图结构，进一步通过GraphSAGE设计局部学习模块及Transformer设计全局学习模块。试验结果表明，相比已有的1^st-ChebNet和GraphSAGE方法，本文方法能够结合局部河段组合特征与全局河网形态特征，做出准确的河网模式识别决策，识别精度可达94%。这为实现智能化河网模式识别提供了一种技术途径。

关键词: 河网模式识别, 几何形态知识, GraphSAGE, Transformer

Abstract:

Drainage patterns recognition is essential for analyzing terrain and geomorphology, exploring geological minerals, and transforming river network data across various scales. However, traditional spatial statistical methods based on morphological and geometric features are not robust enough. To overcome this deficiency, graph convolutional methods have emerged as a popular solution. Nevertheless, these methods often focus narrowly on local features, disregarding the crucial global perspective necessary for comprehensive analysis. To address this issue, our study proposes a drainage pattern recognition method supported by graph Transformer. This method incorporates geometric knowledge by constructing river network graph structures using dual graphs. It integrates a GraphSAGE-based local learning module and a Transformer-based global learning module, training the graph Transformer model. Experimental results demonstrate that our method achieves 94% accuracy in accurately recognizing drainage patterns by combining local segment composite features and global river network morphology features. This outperforms the 1^st-ChebNet and GraphSAGE methods, presenting a promising approach for intelligent drainage pattern recognition.

Key words: drainage pattern recognition, geometric knowledge, GraphSAGE, Transformer

中图分类号:

P208

余华飞, 邱天奇, 周哲, 龚冲亚, 肖天元, 杨敏, 艾廷华. 图Transformer支持下的河网模式识别[J]. 测绘学报, 2024, 53(11): 2075-2085.

Huafei YU, Tianqi QIU, Zhe ZHOU, Chongya GONG, Tianyuan XIAO, Min YANG, Tinghua AI. Drainage pattern recognition supported by graph Transformer[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(11): 2075-2085.

图/表 11

图1

图2

图3

图4

图5

表1

图6

图7

图8

表2

图9

参考文献 29

[1]	ZERNITZ E R. Drainage patterns and their significance[J]. The Journal of Geology, 1932, 40(6): 498-521.
[2]	李敏慧, 吴保生, 陈毅. 黄河源区典型河网平面形态特征及影响因素[J]. 地理学报, 2022, 77(11): 2878-2889.
	LI Minhui, WU Baosheng, CHEN Yi. Planform geometry and controlling factors of river networks in the Yellow River source zone[J]. Acta Geographica Sinica, 2022, 77(11): 2878-2889.
[3]	SULAIMAN M S, UDIN W S, KHAN M A, et al. Analysis of drainage pattern and its relationship with gold potential in Batu Melintang, Jeli, Kelantan[C]//Proceedings of the 3rd International Conference on Tropical Resources and Sustainable Sciences. Kelantan: IOP Publishing Ltd, 2021.
[4]	ZHANG Ling, GUILBERT E. Evaluation of river network generalization methods for preserving the drainage pattern[J]. ISPRS International Journal of Geo-Information, 2016, 5(12): 230.
[5]	JUNG K, SHIN J Y, PARK D. A new approach for river network classification based on the beta distribution of tributary junction angles[J]. Journal of Hydrology, 2019, 572: 66-74.
[6]	ZHANG Ling, GUILBERT E. Automatic drainage pattern recognition in river networks[J]. International Journal of Geographical Information Science, 2013, 27(12): 2319-2342.
[7]	PEREIRA-CLAREN A, GIRONÁS J, NIEMANN J D, et al. Planform geometry and relief characterization of drainage networks in high-relief environments: an analysis of Chilean Andean basins[J]. Geomorphology, 2019, 341: 46-64.
[8]	BURR D M, DRUMMOND S A, CARTWRIGHT R, et al. Morphology of fluvial networks on Titan: evidence for structural control[J]. Icarus, 2013, 226(1): 742-759.
[9]	ICHOKU C, CHOROWICZ J. A numerical approach to the analysis and classification of channel network patterns[J]. Water Resources Research, 1994, 30(2): 161-174.
[10]	LU Yao, YANG Kang, LU Xin, et al. Diverse supraglacial drainage patterns on the Devon ICe Cap, Arctic Canada[J]. Journal of Maps, 2020, 16(2): 834-846.
[11]	MEJÍA A I, NIEMANN J D. Identification and characterization of dendritic, parallel, pinnate, rectangular, and trellis networks based on deviations from planform self-similarity[J]. Journal of Geophysical Research (Earth Surface), 2008, 113(F2): F02015.
[12]	ROCHE M, HELLE M, SAXÉN H. Principal component analysis of blast furnace drainage patterns[J]. Processes, 2019, 7(8): 519.
[13]	YU Huafei, AI Tinghua, YANG Min, et al. A recognition method for drainage patterns using a graph convolutional network[J]. International Journal of Applied Earth Observation and Geoinformation, 2022, 107: 102696.
[14]	YU Huafei, AI Tinghua, YANG Min, et al. Automatic segmentation of parallel drainage patterns supported by a graph convolution neural network[J]. Expert Systems with Applications, 2023, 211: 118639.
[15]	DONADIO C, BRESCIA M, RICCARDO A, et al. A novel approach to the classification of terrestrial drainage networks based on deep learning and preliminary results on solar system bodies[J]. Scientific Reports, 2021, 11: 5875.
[16]	LIU Chengyi, ZHAI Renjian, QIAN Haizhong, et al. Identification of drainage patterns using a graph convolutional neural network[J]. Transactions in GIS, 2023, 27(3): 752-776.
[17]	WANG Wenning, YAN Haowen, LU Xiaomin, et al. Drainage pattern recognition method considering local basin shape based on graph neural network[J]. International Journal of Digital Earth, 2023, 16(1): 593-619.
[18]	KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL]. [2023-12-01]. https://arxiv.org/abs/1609.02907v3.
[19]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2017: 6000-6010.
[20]	HU Yingjie, MAI Gengchen, CUNDY C, et al. Geo-knowledge-guided GPT models improve the extraction of location descriptions from disaster-related social media messages[J]. International Journal of Geographical Information Science, 2023, 37(11): 2289-2318.
[21]	陈晓玲, 唐丽玉, 胡颖, 等. 基于ALBERT模型的园林植物知识实体与关系抽取方法[J]. 地球信息科学学报, 2021, 23(7): 1208-1220.
	CHEN Xiaoling, TANG Liyu, HU Ying, et al. Extracting entity and relation of landscape plant's knowledge based on ALBERT model[J]. Journal of Geo-information Science, 2021, 23(7): 1208-1220.
[22]	吴晨昊, 向隆刚, 张叶廷, 等. 基于地理空间感知型表征学习的轨迹相似度计算[J]. 测绘学报, 2023, 52(4): 670-678. DOI:. doi: 10.11947/j.AGCS.2023.20220026
	WU Chenhao, XIANG Longgang, ZHANG Yeting, et al. Geography-aware representation learning for trajectory similarity computation[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(4): 670-678. DOI:. doi: 10.11947/j.AGCS.2023.20220026
[23]	张金雷, 陈奕洁, Panchamy Krishnakumari, 等. 基于注意力机制的城市轨道交通网络级多步短时客流时空综合预测模型[J]. 地球信息科学学报, 2023, 25(4): 698-713.
	ZHANG Jinlei, CHEN Yijie, PANCHAMY K, et al. Attention-based multi-step short-term passenger flow spatial-temporal integrated prediction model in URT systems[J]. Journal of Geo-information Science, 2023, 25(4): 698-713.
[24]	廖钊宏, 张依晨, 杨飚, 等. 基于Swin Transformer-CNN的单目遥感影像高程估计方法及其在公路建设场景中的应用[J]. 测绘学报, 2024, 53(2): 344-352. DOI:. doi: 10.11947/j.AGCS.2024.20220607
	LIAO Zhaohong, ZHANG Yichen, YANG Biao, et al. Monocular height estimation method of remote sensing image based on Swin Transformer-CNN and its application in highway road construction sites[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(2): 344-352. DOI:. doi: 10.11947/j.AGCS.2024.20220607
[25]	林云浩, 王艳军, 李少春, 等. 一种耦合DeepLab与Transformer的农作物种植类型遥感精细分类方法[J]. 测绘学报, 2024, 53(2): 353-366. DOI:. doi: 10.11947/j.AGCS.2024.20220692
	LIN Yunhao, WANG Yanjun, LI Shaochun, et al. A coupled DeepLab and Transformer approach for fine classification of crop cultivation types in remote sensing[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(2): 353-366. DOI:. doi: 10.11947/j.AGCS.2024.20220692
[26]	STRAHLER A N. Quantitative analysis of watershed geomorphology[J]. Eos, Transactions American Geophysical Union, 1957, 38(6): 913-920.
[27]	DUCHÊNE C, BARD S, BARILLOT X, et al. Quantitative and qualitative description of building orientation[EB/OL]. [2023-12-01]. https://www.researchgate.net/publication/228982283_Quantitative_and_qualitative_description_of_building_orientation.
[28]	HAMILTON W L, YING R, LESKOVEC J, et al. Inductive representation learning on large graphs[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2017: 1025-1035.
[29]	KONG Bo, AI Tinghua, ZOU Xinyan, et al. A graph-based neural network approach to integrate multi-source data for urban building function classification[J]. Computers, Environment and Urban Systems, 2024, 110: 102094.

河网模式类型	样本数量	训练样本（约占60%）	测试样本（约占40%）
树枝状	403	253	150
扇状	401	251	150
平行状	410	260	150
骨架状	443	293	150
矩形状	401	251	150

编号	案例	方法	模式概率
编号	案例	方法	树枝状	扇状	平行状	骨架状	矩形状
（a）		1^st-Cheb Net	0.288 7	0.041 6	0.007 6	0.592 2	0.069 9
		GraphSAGE	0.297 9	0.602 7	0.016 9	0.077 6	0.004 9
		图Transformer	0.720 6	0.000 5	0.000 0	0.278 8	0.000 1
（b）		1^st-Cheb Net	0.000 9	0.170 5	0.828 5	0.000 1	0.000 0
		GraphSAGE	0.112 9	0.862 9	0.024 0	0.000 2	0.000 0
		图Transformer	0.001 2	0.927 1	0.071 6	0.000 1	0.000 0
（c）		1^st-Cheb Net	0.065 5	0.797 6	0.136 8	0.000 1	0.000 0
		GraphSAGE	0.004 7	0.539 6	0.455 6	0.000 0	0.000 0
		图Transformer	0.006 7	0.170 9	0.800 8	0.003 8	0.017 8
（d）		1^st-Cheb Net	0.001 4	0.095 5	0.851 3	0.051 6	0.000 1
		GraphSAGE	0.007 6	0.002 9	0.022 4	0.961 1	0.005 9
		图Transformer	0.000 9	0.000 0	0.000 3	0.965 2	0.033 6
（e）		1^st-Cheb Net	0.749 2	0.216 5	0.003 9	0.003 4	0.027 0
		GraphSAGE	0.004 2	0.000 0	0.000 3	0.795 9	0.199 6
		图Transformer	0.234 5	0.000 1	0.013 1	0.007 4	0.744 8