基于面向对象分类方法的遥感影像变化检测

Object-oriented classification of remote sensing data for change detection

介绍了一种利用GIS数据辅助实现遥感影像变化检测的方法。该方法将面向对象的概念引入遥感影像分类中,使分类的目标不仅可以是影像的像素,也可以是表示一个GIS对象的全部像素集合(像素组)。在影像分类中,为减少工作量和降低分类精度对训练样区选择经验的依赖性,最大似然分类中的训练样区由现有GIS数据对象引导生成。实验表明该方法在针对面状对象进行变化检测时具有良好的效果。

This paper introduces a method regarding the remote sensing data for change detection by using gis database. In this method, the concept of object-oriented has been introduced in, and therefore, the objects of the classification can not only be a single pixel of an image but also pixel sets which represent gis objects. In addition, when remote sensing data are classified by using the supervised maximum likelihood classification method, the training areas can be generated from the gis database to reduce the workload and avoid the dependence on operator＇s experiences. Experiments has demonstrated that this method is effective on detecting the change of area objects.