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：2010
卷号：XXXVIII - Part 7A
页码：53-58
出版社：Copernicus Publications
摘要：Land cover classification plays a key role for various geo-based applications. Numerous approaches for the classification ofsettlements in remote sensing imagery have been developed. Most of them assume the features of neighbouring image sites to beconditionally independent. Using spatial context information may enhance classification accuracy, because dependencies ofneighbouring areas are taken into account. Conditional Random Fields (CRF) have become popular in the field of patternrecognition for incorporating contextual information because of their ability to model dependencies not only between the class labelsofneighbouring image sites, but also between the labels and the image features. In this work we investigate the potential of CRF forthe classification of settlements in high resolution satellite imagery. To highlight the power of CRF, tests were carried outusing onlya minimum set of features and a simple model of context. Experiments were performed on an Ikonos scene of a rural area inGermany. In our experiments, completeness and correctness values of 90% and better could be achieved, the CRF approach wasclearly outperforming a standard Maximum-Likelihood-classification based on the same set of features.
关键词：Conditional Random Fields; contextual information; classification; satellite imagery; urban area
