期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII Part B7
页码：679-684
出版社：Copernicus Publications
摘要：The hybrid classifier is presented in combination of expert system and object oriented approach, for which increased information is added for classification and improve the accuracy. Instead of the original image bands, derived data are prepared for classification, which contains physical meaning and clear separation of recognition and assessment of object classification, the variables of NDVI, seasonal change vegetation index, vegetation brightness index, hard surface brightness index, moisture stress index are derived from raw image data. The 17 classes of land cover are possible for classification and make high overall accuracy of 86%, the classes of evergreen needle leaf, evergreen shrub, sands/beach, lake/reservoir get high accuracy in classification, the classes of grass, grass dominated urban vegetation, bare land/rocks have poor classification, the influence to classification is the factors of DEM, slope, shadow. Expert system based on object orient approach has many advantages of multi-spectral cluster separation and object analysis for land cover classification, especially for more than 10 land cover classes. Multiple variables are better solution in criteria setup than single variable and reduce the hierarchical levels
关键词：Expert classification; Object-oriented classifier; land cover; decision tree; segmentation
