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：Depending on the spatial resolution, the appearance of roads differs in images. In high resolution aerial images a road might be distinguishable as an area with visible road markings, while in a satellite image of low resolution, roads appear as lines and their network character becomes important. The design of the object model for the extraction of roads therefore has to be influenced by the resolution of the available imagery. In this paper we present results of a concept to automatically adapt road models for high resolution images to models suitable for any other lower resolution images with the same spectral characteristics. The road model is formulated as a semantic net, which ensures an explicit representation of objects. Starting from the manually created semantic net for high resolution images and the given target scale, the road model is first automatically decomposed into object parts with similar scale behaviour. The representation of the object parts in the coarser scale is then automatically predicted by scale change models, which are generated by deploying analytical as well as simulation procedures. After scale adaptation of the decomposed object parts, they are fused back to a complete object model in form of a semantic net, which is suitable for road extraction in images of the lower target resolution. The automatically created semantic nets for different scales are used for automatic object extraction. Using an object model for road extraction in urban areas, the developed methods were successfully tested. The presented methodology facilitates the creation of new object models for automatic object extraction in lower resolution images by adaptation, and therefore avoids redundant work
