摘要
山区土地覆盖分类经常受到地形效应及太阳高度角的影响,使用纯粹的基于光谱特征的分类方法很难取得较高的分类精度。本文以从DEM得出的高程、坡度、粗糙度三个地形数据集作为山区土地利用覆盖分类的辅助数据,使用最大似然分类器和基于Back Propagation算法的多层前馈型神经网络分类器分别对上述由光谱数据及地形辅助数据叠和生成的多波段影像进行分类试验,结果显示地形数据辅助下分类结果的精度较原始影像有一定程度的提高。
Classification of mountainous land cover often affected by the differences of topographic effect and height angle of sun, and it is difficult to achieve high accuracy of classification using spectral feature classifier. In the paper, such three topographic data sets as elevation, slope and roughness derived from DEM are used as auxiliary data of classification for mountainous land cov- er, the maximum likelihood classifier and multilayer feed forward network based on back propagation algorithm are employed to do classification experiments on multi-band images developed by superposition of spectral data and auxiliary topographic data. The result indicates that classification aided by topographic data can, in a certain extent, improve the accuracy.
出处
《山东科技大学学报（自然科学版）》
CAS
2007年第1期38-41,共4页
Journal of Shandong University of Science and Technology(Natural Science)
基金
国家高技术研究发展计划(863计划)"规模化高效率土地资源遥感监测业务运行系统"(2003AA131010)资助
关键词
DEM
人工神经网络
BP
遥感影像分类
digital elevation model（DEM）
artificial neural network
back propagation （BP）
classification of remote sensing data
