欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
高分辨率影像包含了丰富的空间信息,这使得基于像素的传统分类方法的分类精度受到局限。基于此,本文对面向对象的分类方法进行了探讨。首先,利用图像的光谱和形状因子对图像进行合理的分割。然后,建立决策树分类判别知识库,将对象归属到某一类上并进行分类。结果显示,面向对象方法的分类精度较传统分类方法有了很大程度的提高,这为通过建立决策树知识库对地物光谱混杂的城区分类提供了一种有益的尝试。
英文摘要：
High resolution remote sensing image contains rich spatial information, for which pixel-based traditional classification methods can＇t satisfy the accuracy requirements. Accounting for this request, object-oriented image analysis method is presented in this paper. Firstly, the image is properly segmented using its spectral and shape factors. Then the distinguishing knowledge base of decision trees is built and objects are assigned to some classes. The results show that the object-oriented approach gives more accurate results than those achieved by traditional classification algorithms, and that it can provide a useful attempt to urban classification of hybrid object spectrum by building the distinguishing knowledge base for decision trees.
同期刊论文项目
同项目期刊论文
期刊信息
位置：