| 图卷积网络在道路网选取中的应用 |
| |
| 引用本文： | 张康,郑静,沈婕,马劲松.图卷积网络在道路网选取中的应用[J].测绘科学,2021,46(2):165-170,177. |
| |
| 作者姓名： | 张康 郑静 沈婕 马劲松 |
| |
| 作者单位： | 南京大学地理信息科学系,南京210023;南京师范大学虚拟地理环境教育部重点实验室,南京210046;南京师范大学地理科学学院,南京210046;江苏省地理信息资源开发与利用协同创新中心,南京210023;南京大学地理信息科学系,南京210023;江苏省地理信息资源开发与利用协同创新中心,南京210023 |
| |
| 基金项目： | 国家自然科学基金项目(41871371)。 |
| |
| 摘 要： | 针对现有制图综合中的道路网自动选取方法不能有效地利用道路网的空间特征问题,该文把道路网抽象为图结构,提出了使用图卷积网络来进行道路网的自动选取,并比较分析了不同的图卷积网络在道路网选取中的适用性。结果表明,图卷积网络可以通过多层卷积来自动提取不同局部范围的空间特征,从而减少空间特征的人工构建,相比传统的多层感知机(MLP)等人工智能选取方法,具有更高的选取精度。对于不同的图卷积网络模型,使用最大池化聚合的GraphSAGE获得了最优的性能。
|
| 关 键 词： | 图卷积网络 道路网 自动选取 制图综合 |
Application of the graph convolution network in the selection of road network |
| |
| ZHANG Kang,ZHENG Jing,SHEN Jie,MA Jinsong.Application of the graph convolution network in the selection of road network[J].Science of Surveying and Mapping,2021,46(2):165-170,177. |
| |
| Authors: | ZHANG Kang ZHENG Jing SHEN Jie MA Jinsong |
| |
| Institution: | (Department of Geographic Information Science,Nanjing University,Nanjing 210023,China;Key Laboratory of Virtual Geographic Environment of Ministry of Education,Nanjing Normal University,Nanjing 210046,China;Institute of Geographic Science,Nanjing Normal University,Nanjing 210046,China;Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application,Nanjing 210023,China) |
| |
| Abstract: | In view of the problem that the existing automatic road network selection methods in cartographic generalization cannot effectively utilize the spatial characteristics of road network,this paper abstracts the road network into a graph structure,and proposes the use of graph convolution networks(GCNs)to automatically select the road network,compares and analyzes the applicability of different graph convolutional networks in the selection of road networks.The results show that the graph convolution networks can automatically extract spatial features of different local ranges through multi-layer convolution,thus reducing the artificial construction of spatial features,and has higher selection accuracy than traditional multi-layer perceptron(MLP)and other artificial intelligence selection methods.For different graph convolutional network models,the best performance is achieved using the max-pooling aggregated GraphSAGE. |
| |
| Keywords: | graph convolution networks(GCNs) road network automatic selection cartographic generalization |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
