论文部分内容阅读
利用ALOS数据,在Definiens Developer 7软件中用分形网络演化法(FNEA)进行多级分割,获取影像对象。综合运用对象的光谱、空间特征和不同层对象之间的关系,提取了湖北省洪湖市试验区土地覆盖与土地利用信息。最后,用一种基于单层分割的面向对象分类方法和基于像素的最大似然法与这种基于多级分割的面向对象分类方法进行了对比分析。结果表明,基于多级分割的面向对象分类方法,不仅克服了基于像素的最大似然法出现的“椒盐”现象,在分类精度上较这两种分类方法也有大幅度的提高。
Using ALOS data, fractal network evolution (FNEA) is used in Multistage segmentation to obtain image objects in Definiens Developer 7 software. The information of land cover and land use in Honghu experimental area of Hubei Province was extracted by comprehensively using the object spectrum, spatial characteristics and the relationship between objects in different layers. Finally, an object-oriented classification method based on single-layer segmentation and pixel-based maximum likelihood method are compared with this multi-level segmentation based object-oriented classification method. The results show that the object-oriented classification method based on multi-level segmentation not only overcomes the phenomenon of “salt and pepper” appearing in pixel-based maximum likelihood method, but also improves the classification accuracy significantly compared with the two classification methods.