基于对象的QuickBird遥感图像多层次森林分类  

Multi-level Forest Classification of Quickbird Remote Sensing Image Based on Objects

作　　者：陈旭[1,2] 徐佐荣[2] 余世孝[2] 

机构地区：[1]西南林学院计算机信息与科学系,云南昆明650224 [2]中山大学生命科学学院生态学系有害生物控制与资源利用国家重点实验室,广东广州510275

出　　处：《遥感技术与应用》2009年第1期22-26,I0001,共6页Remote Sensing Technology and Application

基　　金：西南林学院重点科研基金项目(110712)资助

摘　　要：随着高空间分辨率、相对低光谱分辨率的新型遥感图像的出现和普及,基于像元、依赖于光谱信息特征的传统遥感分类方法已不能很好地适应新型数据的分类工作。通过应用面向对象的多层次分割分类方法对广东省黑石顶自然保护区的QuickBird遥感图像进行了分类尝试。分类结果显示相对于传统的基于像元的分类方法,基于对象的多层次方法具有更准确的分类结果、更明确的分类边界和更均一的内部同质性。With emerging and prevailing of the new remote sensing images that have traditional classification methods based on pixels and spectral information of images has been not adapted to new data. Using the method of object-oriented and multi-level segmentation technique, This paper classified QuickBird remote sensing images of Heishiding Nature Reserve in Guangdong Province. The results showed that compared to traditional pixel-based classification methods,multi-level classification method based on objects can get the more accurate classification results, the more clearly classification boundary and the more uniform internal homogeneity.

关 键 词：对象 多层次分类 图像分割 遥感 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]