tanbul 2004
' directed to
server runs
building
of
end
st choice is
la stored in
| access the
r, and then
files. After
client end
pice is for
coordinate
ling or add
input the
101ce is for
ables users
IL. model).
| interested
1eters. For
he VRML
;, the roof
e building
? name.
re 3. The
read/write
d generate
serve all
> contents.
ages have
he Java™
a Servlet
) Server v
ug-in. The
hich will
user input
tabase
International Archives of the Photogrammetry. Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
3. URBAN 3D MODEL
3.1 Urban 3D Data Structure and Model
Visualization of 3D urban requires a better 3D urban model.
Different applications may have different data types and
manipulation functions. The geometrical information to be
operated to any targeted urban, generally includes two types of
data: vector data and raster data. An appropriate data model
should not only represents the geometrical information, e.g.,
shapes, lengths, areas, etc. but also implicitly or explicitly
Bodies
describe the topological relationship between geometrical
objects, such as adjacency relations, link relations, positional
relations (Wang ef al., 1998; Zhou et al, 2000; Zlatanova,
2000). For our model, the urban object is understood as three
types of data sets (Figure 4):
(1) Digital terrain model (DTM ),
(2) Original images and orthoimages, in our data
structure, texture images are taken as attributes, they
are stored in an independent database.
(3) Spatial objects, such as buildings, roads, waterways.
Attributes
&|
Regular Poly hedron
Quadrics
3 Point
Attributes
Attributes
Height
Sphere
Attributes
Fig. 4. A 3D urban concept model
The most important spatial objects in urban areas are buildings.
There are four different geometric types of objects (Wang et
al., 1998).
(1) Point objects: which are zero-dimensional objects
that have a position but no spatial extension, e.g.,
power poles, wells, etc.;
(2) Line objects: which are one-dimensional objects that
made by connecting two points, e.g., power lines,
telephone lines, etc.;
(3) Face objects: which are two-dimensional objects
with area and perimeter characteristics, such as
parking lots, grass fields, etc.; and
(4) Body objects: which are three-dimensional objects,
such as buildings, barns, etc.
3.2 Implementation in a Relational Database
The data structure and model are implemented by a relational
database technology (Figure 5). Each type of object, shown
above, is defined as a table. The 2D tables are stored in a form
of rows and columns, in which each row is identified by a key
value. Each row stores the information of one instance of an
entity, while cach column describes an attribute of the entity.
For example, a building table includes Building IDs, Roof IDs,
Wall IDs, and attributes (see Figure 5). The Building ID is an
identification code for a building object. A building is made
from a combination of a roof and a wall. The Roof ID points to
a Roof table. A roof table has two terms, Component ID and
Component Type. It means that a building roof is made from a
Component, whose type is either a regular Polyhedron or a
Quadric. A Quadric is one of the three primitives: a Cylinder, a
Cone and a Sphere. These are three basic types of Quadrics in
the VRML modcl. A regular polyhedron is made from Face
LU
objects. This can be seen in the Polyhedron table, with a Face
ID term, which points to a Face table. A face table has two
terms; they are Texture ID and Point ID. Similarly, they point
to Texture tables and Point tables. Just like Roof objects, a
Wall object is also made from a Component. Its table relation is
just the same as that of the Roof object.
4. IMPLEMENTATION THROUGH REAL DATA
4.1 Data Sets
e Aerial Imagery
The test area is located in downtown Denver, Colorado. In this
experimental field, six aerial images were collected on April
17, 2000 using RC 30 aerial camera with a focal length of
153.022 mm at a flying height of 1650 m above the ground
area. The six aerial photographs are formatted along two flight
strips. The aerial photographs were originally recorded in film
and later scanned into digital form at a pixel resolution of 25
um. The endlap of the images is about 65% along strip, and
sidelap is about 30%. Figure 6 shows one of images, DV1119,
whose center is located in the downtown area where numerous
tall buildings are situated.
The six original images, in combination with DSM, and
exterior orientation parameters, are used for generation of
urban TRUE orthoimage. A detailed description for urban
TRUE orthoimage generation can be found in Zhou et al.
(2003). The generated orthoimage is used for base map, on
which the buildings (VRML model) to be built will
be superimposed.
o2
