[关键词]
[摘要]
高分辨率遥感影像中水系的有效提取对于实现快速、高效、范围广的整体水系监测是必不可少的。本文利用易康软件对研究区的遥感影像进行了多尺度分割,选择最合适的分割尺度,并选择最佳的光谱因子 和形状因子(紧致度 和光滑性 ) 进行分割;然后进行面向对象的分类,选取对象的光谱属性信息Layer Values中的Brightness和对象的几何特征Geometry中的length/width两个特征分别采用最邻近分类法和隶属度函数法进行分类,将遥感影像分为水系和其他地类两种情形。面向对象的分类结果要比单纯依靠光谱信息的基于像素的分类方法的结果好,形成的分类结果图更符合人的思维方式。通过和二调成果图的对比以及根据分类结果得出的最佳分类结果概率,表明了最邻近分类法在水系分类中更能准确的进行分类。
[Key word]
[Abstract]
Effective extraction of water system in the high-resolution remote sensing image is essential for the fast, high efficiency, wide range monitoring of the overall water system. In this paper, the study area is made the multi-scale segmentation of remote sensing image by eCognition software, it selects the most appropriate spectral factor , the best shape factor and ) for the segmentation; and then procees to the classification of object-oriented. It chooses Brightness in Layer Values of the object’s Spectral information and length/width in the object’s Geometry for the classification. The classification is made through the nearest neighbor classification method and the membership function classification method. Remote sensing image is classified into water system and other types of land. The classification result of object-oriented is better than the pixel-based classification relying solely on the spectral information. Classification result is more consistent with the way of human thinking. Through the comparison of the land-use classification result map and the best classification result computed in the software, it concludes that nearest neighbor classification is more accurate in the water system classification.
[中图分类号]
[基金项目]