基于知识规则的天然淡水湿地信息提取  

Information Extraction of Natural Freshwater Wetland Based on Knowledge Rules

作　　者：高雷亭 阮仁宗[1] 颜梅春[1] 岁秀珍[1] 傅巧妮[1] 

机构地区：[1]河海大学地球科学与工程学院,江苏南京210098

出　　处：《地理空间信息》2014年第3期33-35,40,共4页Geospatial Information

基　　金：江苏省自然科学基金资助项目(BK2008360);中央高校基本科研业务费专项资金资助项目(2009B12714;2009B11714);江苏省博士后基金资助项目;河海大学人才引进基金资助项目

摘　　要：以洪泽湖天然淡水湿地为研究对象,在光谱特征分析的基础上,选取样本数据,并构建天然淡水湿地信息提取的特征集。利用决策树算法进行数据挖掘和知识规则的生成,并将决策树信息提取结果与最大似然监督分类法进行比较,精度由84.30﹪提高到88.75﹪。In this paper, Hongze Lake was taken as study area and knowledge rules were explored for the extraction of the freshwater wetland in the lake. Classii cation and Regression Tree(CART) algorithm was used to generate the knowledge rules. And then, the rules were applied to the classification of the study area for the extraction of the Hongze freshwater lake wetland. Finally, the results were compared with that of maximum likelihood classii cation results. The study shows that the classii cation accuracy has been increased from 84.30% to 88.75%.

关 键 词：洪泽湖 纹理特征 知识规则 信息提取 分类回归树 

分 类 号：P237.3[天文地球—摄影测量与遥感]