图残差神经网络支持下的建筑物群组模式分类

doi:10.13474/j.cnki.11-2246.2022.0001

测绘通报 ›› 2022, Vol. 0 ›› Issue (1): 1-7.doi: 10.13474/j.cnki.11-2246.2022.0001

• 第八届测绘科学前沿技术论坛获奖论文 • 下一篇

图残差神经网络支持下的建筑物群组模式分类

张自强^1,2,3, 刘涛^1,2,3, 杜萍^1,2,3, 锁旭宏⁴, 杨国林^1,2,3

1. 兰州交通大学测绘与地理信息学院, 甘肃兰州 730070;
2. 地理国情监测技术应用国家地方联合工程研究中心, 甘肃兰州 730070;
3. 甘肃省地理国情监测工程实验室, 甘肃兰州 730070;
4. 中交一航局第二工程有限公司, 山东青岛 266071

收稿日期:2021-05-28 发布日期:2022-02-22
通讯作者: 刘涛。E-mail:ltaochina@foxmail.com
作者简介:张自强(1995-),男,硕士生,主要从事制图综合研究。E-mail:renzailvtu702@163.com
基金资助:
国家自然科学基金（41761088；42061060）；轨道交通工程信息化国家重点实验室（中铁一院）开放课题（SKLKZ21-01）；兰州交通大学天佑创新团队（TY202001）；兰州交通大学优秀平台支持（201806）

Classification of building group patterns using graph residual neural network

ZHANG Ziqiang^1,2,3, LIU Tao^1,2,3, DU Ping^1,2,3, SUO Xuhong⁴, YANG Guolin^1,2,3

1. Faculty of Geomatics, Lanzhou Jiaotong University, Lanzhou 730070, China;
2. National-Local Joint Engineering Research Center of Technologie and Applications for National Geographic State Monitoring, Lanzhou 730070, China;
3. Gansu Provincial Engineering Laboratory for National Geographic State Monitoring, Lanzhou 730070, China;
4. NO.2 Engineering Co., Ltd., CCCC First Harbor Engineering Co., Ltd., Qingdao 266071, China

Received:2021-05-28 Published:2022-02-22

摘要/Abstract

摘要： 建筑物作为城市中的重要地物，分析其群组模式对地图综合、导航定位、市政规划等具有重要作用。建筑物群组模式分析目前主要有基于规则的方法和基于机器学习的方法两种。基于规则的方法和基于传统机器学习分类器的方法均需要大量的人工处理过程。近年来兴起的深度学习特别是图卷积神经网络前期无需人工处理，因此提高了建筑物群组模式分析的自动化程度。传统的图卷积神经网络模型在训练深层网络时易出现退化问题，提取深层特征困难。为解决此问题，本文引入了图残差神经网络模型用于建筑物群组的模式分类。首先使用道路和河流等作为约束条件，利用K-means方法对建筑物进行聚类；然后根据Bertin视觉变量计算对应的建筑物特征指标，在每个建筑物群组中以建筑物质心为节点，连接节点的最小生成树作为边，构建建筑物群组图结构；最后将得到的图结构数据输入图残差神经网络进行训练，得到规则和不规则两种建筑物群组模式。试验结果表明，该模型较好地解决了传统图卷积神经网络模型的退化问题，并取得了更高的精度。

关键词: 建筑物群组, 模式分类, GResNet模型, 机器学习, 深度学习

Abstract: Buildings are important features in the city, the analysis of building group patterns is of great significance in many fields such as map generalization, navigation, municipal planning and so on. Traditional methods for the recognition of building group patterns can be roughly divided into two categories:rule-based methods and machine-learning methods which require a lot of manual processing. In recent years, deep learning, especially the graph convolution neural network's emerging that does not require prior manual processing and can improve the automation degree of the analysis of building group patterns. Traditional graph convolutional neural network model is prone to degradation when training deep networks, which makes it difficult to extract deep features. To solve this problem, a graph residual neural network (GResNet) model is proposed for the classification of building group patterns. Firstly, the roads and rivers are used as constraints, and the K-means method is used to cluster the buildings. Secondly, many indices are used to compute Bertin's visual variables. In each building group, the centroids of the buildings are taken as nodes, and the minimum spanning tree is used to generate edges connecting nodes, after that, the graph representation for building group is constructed. Finally, the building graphs are taken as the input of the proposed GResNet model, and two building group patterns are obtained, namely, regular groups and irregular groups. Experiment results confirm that the proposed model can solve the degradation problem of the traditional graph convolutional neural network model, and obtain higher accuracy.

Key words: building groups, pattern classification, GResNet model, machine learning, deep learning

中图分类号:

P237

张自强, 刘涛, 杜萍, 锁旭宏, 杨国林. 图残差神经网络支持下的建筑物群组模式分类[J]. 测绘通报, 2022, 0(1): 1-7.

ZHANG Ziqiang, LIU Tao, DU Ping, SUO Xuhong, YANG Guolin. Classification of building group patterns using graph residual neural network[J]. Bulletin of Surveying and Mapping, 2022, 0(1): 1-7.

参考文献

[1] BRASSEL K E, WEIBEL R. A review and conceptual framework of automated map generalization[J]. International Journal of Geographical Information Systems, 1988, 2(3):229-244.
[2] CARMONA M, CARMONA R, HEATH T, et al. Public places, urban spaces:the dimensions of urban design[M].[S.l.]:Architectural Press,2003.
[3] WU B, YU B L, WU Q S, et al. An extended minimum spanning tree method for characterizing local urban patterns[J]. International Journal of Geographical Information Science, 2018, 32(3):450-475.
[4] ZHANG X, STOTER J, AI T H, et al. Automated evaluation of building alignments in generalized maps[J]. International Journal of Geographical Information Science, 2013, 27(8):1550-1571.
[5] WURM M, SCHMITT A, TAUBENBÖCK H. Building types' classification using shape-based features and linear discriminant functions[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(5):1901-1912.
[6] HE X J, ZHANG X C, XIN Q C. Recognition of building group patterns in topographic maps based on graph partitioning and random forest[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 136:26-40.
[7] ZHANG L Q, DENG H, CHEN D, et al. A spatial cognition-based urban building clustering approach and its applications[J]. International Journal of Geographical Information Science, 2013, 27(4):721-740.
[8] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553):436-444.
[9] BRONSTEIN M M, BRUNA J, LECUN Y, et al. Geometric deep learning:going beyond euclidean data[J]. IEEE Signal Processing Magazine, 2017, 34(4):18-42.
[10] ZHAO R, AI T H, YU W H, et al. Recognition of building group patterns using graph convolutional network[J]. Cartography and Geographic Information Science, 2020, 47(5):400-417.
[11] YAN X F, AI T J, YANG M, et al. Graph convolutional autoencoder model for the shape coding and cognition of buildings in maps[J]. International Journal of Geographical Information Science, 2021, 35(3):490-512.
[12] YAN X F, AI T H, YANG M, et al. A graph convolutional neural network for classification of building patterns using spatial vector data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2019, 150:259-273.
[13] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas:IEEE,2016:770-778.
[14] DEFFERRARD M, BRESSON X, VANDERGHEYNST P. Convolutional neural networks on graphs with fast localized spectral filtering[J].neural information processing systems,2016:3844-3852.
[15] DU S H, LUO L Q, CAO K, et al. Extracting building patterns with multilevel graph partition and building grouping[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 122:81-96.
[16] DENG Min, TANG Jianbo, LIU Qiliang, et al. Recognizing building groups for generalization:a comparative study[J]. Cartography and Geographic Information Science, 2018, 45(3):187-204.
[17] BIEDERMAN I. Recognition-by-components:a theory of human image understanding[J]. Psychological Review, 1987, 94(2):115-147.
[18] PEURA M, IIVARINEN J. Efficiency of simple shape[J].Advances in Visual Form Analysis,1997:443-451.
[19] BASARANER M, CETINKAYA S. Performance of shape indices and classification schemes for characterising perceptual shape complexity of building footprints in GIS[J]. International Journal of Geographical Information Science, 2017, 31(10):1952-1977.

图残差神经网络支持下的建筑物群组模式分类

Classification of building group patterns using graph residual neural network

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	孔瑞瑶, 谢涛, 马明, 孔瑞林. CatBoost模型在水深反演中的应用[J]. 测绘通报, 2022, 0(7): 33-37.
[2]	温雨笑, 吕杰, 马庆勋, 张鹏, 徐汝岭. 高光谱和LiDAR联合反演森林生物量研究[J]. 测绘通报, 2022, 0(7): 38-42.
[3]	刘立, 董先敏, 刘娟, 文学虎. 人机融合智能的遥感解译生产新方法[J]. 测绘通报, 2022, 0(7): 118-123,137.
[4]	蔡柔丹. 一种基于用户异步轨迹的身份识别智能方法[J]. 测绘通报, 2022, 0(7): 158-162,167.
[5]	滕文鑫, 张建辰, 邵杰. 集成时空邻近与卷积网络车道级高精度定位算法[J]. 测绘通报, 2022, 0(6): 1-5,61.
[6]	张晋赫, 秦育罗, 张在岩, 宋伟东, 朱洪波. 复杂场景下农村道路裂缝分割方法[J]. 测绘通报, 2022, 0(5): 74-78,88.
[7]	师芸, 石龙龙, 牛敏杰, 赵侃. 基于单阶段实例分割网络的黄土滑坡多任务自动识别[J]. 测绘通报, 2022, 0(4): 26-31.
[8]	王雪, 梁珂, 隋立春, 钟棉卿, 朱剑锋. 膨胀卷积与金字塔表达的深度学习模型用于农村建筑物提取[J]. 测绘通报, 2022, 0(4): 61-65.
[9]	张驰, 白志辉, 李亮, 陈冉丽, 吴侃. 利用三维激光扫描技术监测矿区桥梁结构形变[J]. 测绘通报, 2022, 0(4): 122-129.
[10]	高鸣, 周鑫鑫, 刘琦, 杨光迪, 吴长彬. 深度学习网络支持下的农房侵占耕地自动化监测[J]. 测绘通报, 2022, 0(3): 47-53.
[11]	王刚, 丁华祥. Sentinel-2A数据支持下的雷州半岛植被类型识别[J]. 测绘通报, 2022, 0(3): 76-82.
[12]	孙琴琴, 蔡国印, 杨柳忠, 张宁, 杜明义. 城市尺度典型地表要素综合提取方法研究[J]. 测绘通报, 2022, 0(2): 62-66.
[13]	车子杰, 高飞, 吴兆福, 李振轩. 基于改进U-Net网络的多源遥感影像洪涝灾害信息提取与变化分析[J]. 测绘通报, 2022, 0(1): 26-32.
[14]	程瑞, 高建, 邢强, 孙中昶. 一种优化的Faster R-CNN小目标检测方法[J]. 测绘通报, 2021, 0(9): 21-27.
[15]	淳锦, 张新长, 郭海京, 张建国, 金诗程. 基于多源信息与深度学习特征提取的人口空间抽样方法[J]. 测绘通报, 2021, 0(8): 42-47.