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：2005
卷号：XXXVI-8/W27
出版社：Copernicus Publications
摘要：Commercially available high-resolution satellite imagery from sensors such as IKONOS and QuickBird are important data sources for a variety of urban area applications including infrastructure feature extraction and land cover mapping. Land cover maps from medium and high-resolution imagery are typically generated through supervised spectral classification of multispectral imagery. Supervised classification algorithms require training data as input and are thus semi-automated approaches. However, by automating the generation of training data, these supervised classifiers can be utilized in a fully automated, or self-supervised fashion to perform urban land cover classification. In this paper, we present a self-supervised approach for fully automated urban land cover classification of high-resolution satellite imagery. Automated feature extraction techniques are utilized to generate training data that are then input into supervised classification algorithms, thereby producing a self-supervised urban land cover classifier. These feature extraction techniques do not seek to extract all features present in the imagery. Instead, they are used to identify very high confidence instances of the different urban land cover classes. In this way, we limit the amount of incorrect training data that is input into the classifier. Because labeled training data is generated internally by the system, this classification approach is referred to as self-supervised. Self-supervised classification systems differ from unsupervised classifiers in that unsupervised classifiers output an unlabeled classification, requiring further analysis to determine the class labels, whereas the output of a self-supervised classifier is a labeled classification. Initial test results indicate that the overall accuracy of the self-supervised classification is 87-93%. There is only a 2% increase in overall accuracy when manually supervised classification is performed on the same test site
关键词：Self-supervised classification; high-resolution satellite imagery; urban land cover; feature extraction
