基于深度学习的中文扫描地图注记提取方法  

A method of chinese scanned map annotation extraction based on deep learning

作　　者：饶珣 王加胜[2] Rao Xun;Wang Jiasheng(School of Information Science and Technology,Yunnan Normal University,Kunming 650500,China;Faculty of Geography,Yunnan Normal University,Kunming 650500,China)

机构地区：[1]云南师范大学信息学院,云南昆明650500 [2]云南师范大学地理学部,云南昆明650500

出　　处：《无线互联科技》2023年第4期129-132,共4页Wireless Internet Technology

摘　　要：地图注记是地图的核心内容之一,地图注记提取是计算机智能读取扫描地图信息的重要步骤。中文扫描地图注记由于字符多样、线条复杂、方向多变,给中文扫描地图注记矢量化带来了困难,目前少有研究。文章提出了一种基于深度学习的中文扫描地图注记提取方法,分为注记特征提取和空间重建两个部分。结果显示,该方法能够实现地图中文注记的检测、识别和空间重建。Map annotation is one of the core contents of map,and the vectorization of map annotation is an important step for computer to intelligently read and scan map information.The vectorization of Chinese scanned map annotation is difficult due to its diverse characters,complex lines and changeable directions.In this paper,we propose a vectorization method of Chinese scanned map annotation based on deep learning,which is divided into two parts:annotation feature extraction and spatial reconstruction.

关 键 词：地图注记 深度学习 矢量化 特征提取 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]