土地利用现状图符号的自动提取与识别.pdfVIP

DOI ：CNKI:11-4415/P1814.018 网络出版时间：2010-11-19 18:14 网络出版地址：/kcms/detail/11.4415.p1814.018.html 土地利用现状图扫描符号的自动提取与识别 ① ② 农 宇 ，陈 飞 （①武汉大学资源与环境科学学院，武汉 430079 ；②广州市城市规划勘测设计研究院，广州 510060 ） 【摘 要】土地利用现状图扫描符号的自动提取能节省人力财力物力。本文采用地图代数和人工神 经网络的方法进行扫描现状图符号的自动提取和识别。首先根据图例建立用于特征匹配的符号模板 库，然后使用加壳变换和蜕皮变换进行点状符号和线状符号的分离，并使用Canny 算子平滑符号， 再通过人工神经网络的训练进行色彩聚类而实现符号分割。定义多重相似度，实现符号的预分类识 别，在此基础上计算加权距离再次识别。最后采用加壳变换连接断点并提取骨架线对这些断线进行 连接，形成完整的线状符号。实验表明，本方法的精确度达到90% 以上。 【关键词】土地利用现状图；扫描符号；符号分割；符号识别；地图代数 【中图分类号】P283.7 【文献标识码】A 【文章编号】1009-2307 （2011）02- - Auto extraction and identification for scanned symbols of land use map Abstract: Auto extraction and identification for Scanned Symbols of land use map contributes to a more efficient work. This paper discussed the automation in a map algebra way. An artificial neural network (ANN) method was also applied for sample training. Point and line symbols were firstly extracted by distance transform based on the symbol template, then edges were smoothed by Canny operator. After ANN training for clustering those symbols were extracted to make preliminarily classification. Disconnections in the skeleton line were performed distance transform again at last for completing line symbols. The result showed an accuracy beyond 90%. Key words: land use map; scanned symbols; symbol extraction; symbol identification; map algebra ① ② NONG Yu , CHEN Fei (①School of Resource and Environmental Science, Wuhan University, Wuhan 430079, China; ②Guangzhou Urban Plan