基于地图内容特征的网络地图服务分类方法研究  

Research on Network Map Service Classification Method Based on Map Content Features

作　　者：陈卓 孙建军 CHEN Zhuo;SUN Jianjun(Tianjin Survey Design Institute Group Co.,Ltd.,Tianjin 300191,China)

机构地区：[1]天津市勘察设计院集团有限公司,天津300191

出　　处：《测绘与空间地理信息》2023年第6期169-171,174,共4页Geomatics & Spatial Information Technology

摘　　要：随着地理信息资源共享的全球化,能够高速且精确地检索用户需要的WMS成为当前研究的重要方向。目前,研究大多以文本信息为基础,结合空间关系和文本匹配进行资源检索,缺少对地图视觉特征的分析,忽略了地图内容特征所表现出的地理空间现象。为此,本文设计并实现了一种基于图像内容的WMS图层分类方法,尝试从颜色、形状、纹理3个角度对WMS图层特点进行描述,分别使用经典特征提取结合SVM机器学习和基于CNN(VGG-19)的迁移学习两种方式对WMS图层进行分类,对实验结果进行比较及适用性验证。结果表明,本文提出的方法能够在WMS图层数据集中取得较高的识别准确度,对于顾及内容的网络地理信息资源检索具有一定的指导意义。Research on network map service classification method based on map content features with the globalization of geographic information resource sharing,it has become an important research direction to retrieve WMS needed by users quickly and accurately.At present,most studies are based on text information,combined with spatial relationship and text matching for resource retrieval,lack of analysis of map visual features,and ignore the geospatial phenomenon shown by map content features.Therefore,this paper designs and implements a WMS layer classification method based on image contents,tries to describe the characteristics of WMS layers from the perspectives of color,shape and texture,classifies WMS layers by using classical feature extraction combined with SVM machine learning and migration learning based on CNN(VGG-19),compares the experimental results and verifies their applicability.The results show that the method proposed in this paper can achieve high recognition accuracy in WMS layer data set,and has certain guiding significance for network geographic information resource retrieval considering contents.

关 键 词：网络地图服务 图像特征提取 支持向量机 迁移学习 

分 类 号：P283.7[天文地球—地图制图学与地理信息工程] TP391.41[天文地球—测绘科学与技术]