中文摘要：

选取钱塘江中游地区约348 km2为实验区,综合归一化植被指数、纹理信息和数字地形模型（DEM）派生的高程、坡度等辅助数据,对SPOT5影像的光谱特征进行扩展,建立基于C5.0算法的模型,实现对土地利用信息的自动提取,并将分类结果与基于传统像元的最大似然分类结果作比较。结果表明,训练样本的有效性和辅助特征数据的参与可排除干扰信息;随着样点数量的增加,分类精度提高;决策树向决策规则的转化,能够在保证精度的基础上使规则更易理解;利用C5.0算法的总精度达到94.68%,较最大似然分类法提高了7.37%,有效实现了高精度分类,是保证钱塘江流域地区土地利用遥感信息快速准确提取的方法之一。

英文摘要：

Based on the 348 km2experimental area located at the middle part of Qiantang River,the research integrated NDVI,texture features,elevation and slope features generated from DEM and other spatial data,extended SPOT5 image’s spectrum features,and automatically extracted land-use information by C5.0 Algorithm.Comparing to the maximum likelihood classification,the result showed the validity of training samples and usage of assistant feature data could eliminate the disturbing information.When the number of sample points increased,the accuracy of classification could be improved subsequently.The rules were easy to understand while the accuracy was the same,when decision tree changed to the decision rules.By using C5.0 Algorithm,the total accuracy could reach 94.68%,which was 7.37% higher than maximum likelihood classification.C5.0 Algorithm achieved high accuracy classification,which is one of the quick and accurate methods to extract land-use information on Qiantang River drainage area.