(TUBITAK CAREER Award)
We developed an urban visualization framework for the population and visualization of complex urban environments (see Figure 1). To this end, we develop a novel data structure for the storage and representation of buildings in an urban environment, called slicewise representation, which reduces the computational and storage costs of the modeling, visualization and rendering of urban environments; an occlusion culling algorithm, which speeds up the visualization of city models during walkthrough in these environments. The slice-wise representation is based on the observation that the visible parts of the buildings in a typical urban walkthrough are mostly in one of the following three cases (see Figures 2 and 3). Our occlusion culling algorithm is a conservative algorithm based on occluder shrinking. Figure 4 illustrates the shrinking method we used applied on a three dimensional object.
Figure 1: Overview of the urban visualization framework: in the first phase, we read the scene and calculate the bounding boxes of objects. Next, we apply uniform subdivision to each object. Then, we cluster the cells of the uniform subdivision into slices. After creating the shrunk versions of the objects, these slices are checked for occlusion and a tight visible set is determined for each grid location. The phases in dashed blocks are performed in the preprocessing phase. The view-frustum culling (VFC) is also done during navigation.
Figure 2: Visibility forms during urban navigation: (a) L-shaped form; (b) vertical rectangular form; (c) horizontal rectangular form. In each part of the figure, the visible part of the occludee is the green transparent area.
Figure 3: Illustration of the visibility forms during a typical urban walkthrough.
Figure 4: Illustration of the proposed occluder shrinking approach applied to a 3D object.
We propose techniques for the monoscopic and stereoscopic visualization of urban environments. To speed up the stereoscopic visualization, we propose to generate the view for one eye using the view for the other, instead of generating the two views separately. We also exploited the Graphics Processor Unit and the related data structures, such as Vertex Buffer Objects, to speed up the visualization process (see Figure 5).
Figure 5: The VBO data structure used in GPU-based visualization. The object triangles are constructed using the index buffers created in the GPU and accessed as needed for each building and for each slice.
We model the crowds in urban environments using a model-based approach where each agent can show complex behaviors based on its skeletal motion. We use occlusion culling to eliminate the agents that are occluded by the buildings and view frustum culling to eliminate agents that are out of the view frustum for fast rendering of crowds. We also use different levels of details of the agents to render the agents according to the distance from the viewpoint. We also work on approaches for global and local path planning for the behavior of agents in crowds. We model the spreading of the panic behavior throughout the crowd. We propose techniques to simulate the behavior of agents for normal and emergency situations, such as fire, explosion, and terrorist attacks, in outdoor environments (see Figure 6). We also study the effects of personality traits to the crowd behavior for indoor environments, such as museums, exhibition galleries (see Figures 7 and 8).
Figure 6: Normal crowd behavior (left) and emergency crowd behavior during explosion and fire in an outdoor environment.
Figure 7: People with low conscientiousness and agreeableness cause congestion.
Figure 8: Neurotic, nonconscientious, and disagreeable agents (in green suits) display panic behavior.
The developed techniques and the proposed framework can be used in urban planning, military simulations, geographical information systems, tourism, entertainment industry, including the film sector and computer games, and security applications.
April 2005 - April 2010
169,000 YTL (~US$110,000)