A qualitative and quantitative visibility analysis in urban scenes

摘要

In this paper we present a mathematical model of a generic urban scene, which can be used to answer a variety of visibility questions. Aside from its theoretical merit, the mathematical model provides an important analysis tool to optimize urban walkthrough algorithms. We formulate the probability for a given object to be visible from a given viewcell as a function of distance from the viewcell. We address various issues related to the implementation of virtual walkthrough, such as storage requirements, optimal viewcell size and cell-to-cell coherence, from which we derive space-effective data structures. Quantitative simulations verify the validity of our analysis. We simulate visibility in scenes with randomly distributed (Poission) objects as well in pseudo-random (jittered) scenes, and compare these simulation results with our mathematical model.