面向对象的多尺度水体信息提取

以安徽省阜阳市阜南县东南部为研究区,利用LandsatTM数据,基于面向对象的多尺度分类方法,选取合适的分割尺度创建分类对象;通过分析研究区各种地物复杂的光谱响应特征,建立水体信息提取的知识规则集,对研究区水体进行提取研究。研究表明,与传统的利用单波段和NDWI等方法所提取的水体相比,规则推理法在提取水体时可以消除一些影像上其他干扰信息的影响,取得较好的提取效果和精度。除此之外,该方法简单易行,水体提取的规则集对于用户来说是透明的,易于理解。

Information Extraction of Object-oriented Multi-scale Water Bodies

In the paper, the south-eastern part of Funan country was taken as study area. Land sat TM is used for the extraction of water bodies. After the analysis of objects, the optimal scale was found for the segmentation of remotely sensed data and hierarchical objects were created by using Multi-scale segmentation algorithm. Based on the analysis of complicated spectral features of objects in the study area, knowledge rule set was generated found for the classification of objects in eCognition. Compared with the results of water bodies extracted by using the traditional single band techniques and NDWI technique, the method of rule inferring can achieve high accuracy in the extraction of water bodies. Moreover, the method is simple and easier to apply, and apparent to the user.