You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

The 3rd ISPRS Workshop on Dynamic and Multi-Dimensional GIS & the 10th Annual Conference of CPGIS on Geoinformatics
Chen, Jun

ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS", Bangkok, May 23-25, 2001
In the cybercity, the most important components are the
buildings and the municipal facilities. So the 3D reconstruction of
buildings and facilities is a significant work for cybercity.
According to the LOD (level of detail) concept, CCGIS provides
three different 3D modeling schemes for different data sources,
as shown in Fig. 2. The authors agreed with the views of Woo
Sung Ye (1997), it is better to use the fractal dimensions to
describe the LODs of 3D city models. The larger the dimension,
the higher the level of detail, and vice versa. For instance, point
is 1D and planar line is 2 D, but the DEM surface can be
considered as 2.5D. The outline box of building is also only 2.5D,
if the real exterior shape of building such as the roof is
expressed the model is 2.75D, and the real 3D entity should
express all the details including the exterior and the interior
structure. As shown in Fig. 1, there are usually 4 different 3D
modeling methods for cybercity construction.
(1) The 2.5D virtual landscape can be generated by
overlapping the orthoimage onto the DEM surface.
(2) If designed data such as CAD/3DS/3DMAX is used, the real
characteristics of geometric structure and texture can be
expressed. Such kind of 3D city models can reach more
higher level of detail. Not only the cityscape, but also the
rooms and the furnishings setup can be expressed
(3) Based on the photogrammetry, laser scanning and other
ground surveying means, 3D coded data and real texture
images can be obtained efficiently to describe the status of
cityscape. Such kind of modeling methods are therefor used
broadly for large-scale modelling with multiple scales and
relatively higher detail level (2.75D).
(4) Another simple modelling method is to use the bottom
boundary data of buildings(i.e. the 2D vectors of GIS) and
its height attributes. Since the city model created by this
method is only 2.5D (all the roofs are planar), it is usually
used to express the profile of cityscape with relatively lower
level of detail.
Based on the above multiscale modeling methods, from the
street to each building and even to each room, various details of
city model can be expressed. Because the CAD based methods
and the simple method have been used successfully, the third
method based on the digital photogrammetric workstation (DPW)
is therefore discussed in detail. Aerial image and close-range
image are very important data source for building reconstruction,
especially the DPW provides the most economic and fast
technique for 3D city data collection. In order to simplify the data
collection flow, to reduce the workload and to ensure the
reliability of 3D reconstruction, in CCGIS we recommend the
following scheme.
(1) The city model is classified into different basic classes
(primitive entities) according to the shape characteristics,
such as point, line, simple entity, multi-entity, sphere,
cylinder, convex surface, concave surface and upright
surface, and each voxel is identified by class code and user
code. Arbitrary complex buildings can be composed by
these primitives.
(2) The difficulties of 3D reconstruction (without regard to the
automatic extraction from image) is mainly the express of
special roof shape and general concave surface. But based
on the feature points and feature lines obtained by DPW
and automatic TIN algorithm, CCGIS solves this problem
preferably. Fig. 2 illustrates this kind of multi-scale
1 Ü u
r-i nJ
Figure 2 2.5D model generated from the 2D outline
(b) 2.7D model generated from the 2D roof outline
Figure 2 Geometric models with different detail levels
(3) Complex city model then can be assembled in an interactive
edit environment by using of voxel primitives, and all the
attributes including texture, material and multimedia
description also can be related. If necessary the existing
model can be edited, like to append or to delete any kind of
object (point, line, surface, or body), to modify the geometric
shape of buildings such as to change the planar roof to
herringbone roof.
(4) All the modeling results are stored in a hybrid system of files