When combining a reduction in the number of polygons with texture mapping, usually a reduction by at least 90% of the original size can be achieved with little visual impact. A good overview on mesh simplification techniques is given in (Heckbert and Garland, 1997). To give an example fora mesh simplification, figure 2 shows part of a DSM of our test area which was overall modeled using 607624 triangles. Figure 3 shows the same scene after mesh simplification with only 50000 triangles for the whole test area.Triangle reduction was performed using the POLYWORKS software package (Soucy, 1997).
Although the amount of data can be reduced significantly by these algorithms, many tasks aiming on visualizations or simulations in an urban environment require the further abstraction and interpretation of the surface description. For simulations aiming on the propagation of noise or electromagnetic waves e.g. knowledge about the surface material is a crucial point. Hence trees or buildings have to be represented separately from the terrain surface. The situation is much like in reverse engineering for industrial parts where the ultimate goal is to find a compact parametric representation rather than dense point clouds.