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：2004
卷号：XXXV Part B3
页码：1178-1183
出版社：Copernicus Publications
摘要：An object-based approach is applied in land-cover feature extraction from high-resolution multi-spectral images and laser scanning data in this research. Objects extracted from high-resolution spectral images and laser data may have both classification errors and geometric errors. The classification errors are mainly caused by uncertainty of class definition (fuzzy classes), limited information content in the RS data and uncertainty of the validity of the decision rules for classification. These sources of uncertainty affect the quality of thematic or classification results of the information extraction from RS data, i.e. the thematic uncertainty. On the other hand, there are several types of errors that have an impact on the object location and the spatial extent of objects such as the geometry of object boundaries and the size of objects. These sources cause the spatial uncertainty of the information extracted from RS data. To assess the uncertainty of objects, we have to consider both types of uncertainty at the same time since they are interrelated. In this paper, we will firstly model these two different types of uncertainty separately and then analyse how these two types of uncertainty interact. This analysis will consider several parameters factors such as characteristics used for extracting objects, classification rules, as well as the minimum mapping units. Finally we will propose a unified framework considering both thematic and spatial uncertainty for the evaluation of the uncertainty of the results of information extraction from RS data. An overall uncertainty measure is proposed that is context-related and that will emphasise the type of quality relevant for particular applications. This will be illustrated by a case study using a 1 m resolution multi-spectral image and laser scanning data of an urban area. The experimental results confirm the effectiveness of the proposed uncertainty measures
关键词：Quality; Uncertainty; Object; Extraction; High resolution; Classification; Land Cover; LIDAR
