In the last years more and more applications make use of 3d building models. Automatic reconstruction is essential for the acquisition and updating of the huge amount of data. For many applications it is useful to be able to transform the object representation automatically in order to generalize the geometry. For this task a structurally rich description of the object is needed. We presented a reconstruction procedure which models the structure of facades. This method is based on a facade grammar and a reversible jump Markov Chain Monte Carlo process. For the stochastic process we need prior knowledge about facades. The distribution of facade elements can influence the proposal of rules. Until now we made general assumptions about these distributions. In this paper we analyze facades from a set of facade images and derive prior facade information which support the modelling process.
展开▼
