摘要
为获得高分辨率影像荒漠植被信息的专家知识库,采用多尺度分割技术,探索荒漠植被信息的光谱特征、纹理特征和上下文特征,建立荒漠植被信息与特征之间的对应关系.结果表明,通过多尺度分割,可体现土地覆盖的层次性.同时,在专家知识库的基础上,基于对象分类法的精度相对比较高,总体分类精度为85.96%,Kappa系数为0.81,分类结果没有"椒盐现象",且各种地物图斑边界明确而规整,易于分类后处理.
The information-characteristics correspondence of desert vegetation was built with multiresolution segmentation technique through exploring the features of spectrum,texture,and content of desert vegetation to obtain its expert knowledge base of high resolution image.The result shows that the hierarchy of land cover can be embodied from multi-resolution segmentation.On the basis of expert knowledge base,the object-based classification method is with high precision that the total classification accuracy is 85.96%,and the Kappa coefficient 0.81 with no＂pepper and salt phenomenon＂appeared in classification result,the terrain map spots are clear-cut and in order,which is liable to post classification process.
出处
《宁夏大学学报（自然科学版）》
CAS
2015年第1期66-72,共7页
Journal of Ningxia University(Natural Science Edition)
基金
国家自然科学基金资助项目(41271360)
关键词
基于对象的影像分类
专家知识库
高分辨率影像
荒漠植被
信息提取
object-based image classification
expert knowledge base
high resolution image
desert vegetation
information extraction
