深度学习在地图多尺度表达中的应用分析  

作　　者：李俊 吴长枝[2] 齐晓飞[2] 赵耀 

机构地区：[1]速度时空信息科技股份有限公司,南京210000 [2]西安测绘研究所,西安710054

出　　处：《科技创新与应用》2023年第12期185-188,192,共5页Technology Innovation and Application

摘　　要：传统方法在地图多尺度表达研究中存在地图要素结构化描述困难、算法自适应程度不高和空间关系保持能力有限等问题。深度学习在图像解译和时空预测中的成功应用,给地图多尺度表达提供全新思路。根据地图要素几何类型(线要素、面要素),归纳当前深度学习算法用于提取空间分布模式、模拟制图综合过程的相关研究;分析深度学习网络模型在地图多尺度表达中的应用效果,总结当前研究存在的问题和下一步改进方向。In the research of map multi-scale representation,there are some problems in traditional methods,such as difficulties in structured description of map elements,low degree of algorithm adaptation,limited ability to maintain spatial relations and so on.The successful application of deep learning in image interpretation and spatio-temporal prediction provides a new idea for multi-scale representation of maps.According to the geometric types of map elements(line features,area features),this paper summarizes the relevant research on the current deep learning algorithms for extracting spatial distribution patterns and simulating the process of cartographic generalization,analyzes the application effect of deep learning network model in map multi-scale representation,and summarizes the existing problems in the current research and the direction of further improvement.

关 键 词：深度学习 地图多尺度表达 图卷积 神经网络 道路网 建筑物 

分 类 号：P231[天文地球—摄影测量与遥感]