Visibility Culling for Large Dynamic Crowds within Buildings

Visibility calculation of dynamic scenes is a key issue for real-time rendering of large scale scenes,but a time consuming process,in particular,for the visibility computation of large dynamic crowds.In this paper,we present an efficient visibility culling method for dynamic crowds moving inside buildings.During the preprocessing stage,the space of a building is organized as a binary tree,our method subdivides the entire simulation time period into discrete intervals and assumes that at each time interval,any individual of the crowd remains in the same subspace of one scene node while accounting for his/her movement,therefore can be bound to that node.During the rendering stage,the visibility of a scene node can be used as an over-estimate of the visibility of all individuals bound to that node regarding the specified time interval.Experiments show that our method can efficiently cull invisible individuals of dynamic crowds within complex buildings,thus greatly accelerate the visibility calculation process.