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：2007
卷号：XXXVI-3/W49A
页码：111-116
出版社：Copernicus Publications
摘要：We present a system for automatic building reconstruction, combining the strengths of grammars to generate varying building modelsand the principle of minimum description length [MDL] to evaluate results and to control the search process. The reconstruction processis guided by the level of detail, starting with a coarse level and is stepwise improved towards highly detailed models. On each levelof detail, the corresponding components are identified by prototypes, provided by the building ontology. The matching between inputsurfaces and prototypes is supported by constraints representing relevant topological and geometrical relations between these surfaces.When employing MDL, usually an optimal coding is required. In contrast, we use an asymptotic optimal coding which is easier togenerate since no a priori knowledge is needed. Instead, a larger amount of input data has to be processed to achieve comparable results.
