it 1S
; of
, in
(ion
igh
vital
per
1ips
ruct
the
AR
ting
igh-
GIS
oral
oral
odel
ition
ilize
the
data
x of
arch
vely
“the
cond
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
- Terrain: the digital terrain data including Digital
Elevation Model derived from high-resolution remote
sensing data covered with orthoimages;
> Buildings: including the spatial location and attributes
such as the height of the building and so on;
- Vegetation: including the elevated vegetation just as a
tree or forest, and the ground land use information;
- Time stamp: the time state and its changes
corresponding to the spatial model;
- Objects attributes: the ground objects states description;
And the above factors will be discussed in reconstructing
spatio-temporal model from airborne SAR images.
3. TGIS-SURPPORTED SPATIAL MODEL CELL
3.1 Spatio-temporal model cell conception
A Temporal Geographic Information System (TGIS) is more
truthful in the simulation of real world phenomena because it
includes more real world dimensions of space, time, and
attribute than a static GIS. At the present, in China, land use
changes is very rapid and wide, thus the renewing Land Use
Dynamic Monitoring Database (LUDMD) using remote sensing
imagery is a frequent process. And so, the LUDMD may be
researched as a TGIS.
The object-oriented spatio-temporal model based on space and
time cube pointed by Worboys, 1992, has considered the
changes of space, with the absence of attributes changes. This
paper expands the thought and puts forward the object-oriented
spatio-temporal cell based on real terrain surface segments. The
real terrain surface will be segmented into many facets in
conception, conformed with the ground altitude and the
attributes with time stamp of objects including the space
location distribution, height, state and so on, to form a spatio-
temporal model cell. The whole spatio-temporal model
comprised of many spatio-temporal cells, which can be
reconstructed from high-resolution remote sensing data such as
airborne SAR images and essential GIS data.
€
3.2 Space and time representation
The three elements of spatio-temporal model cell include the
space distribution, time information and object attributes. And
the spatial and temporal changes of the cell are described as a
4.5 dimension variable, that is, space with 3 dimensions, time
with 1 dimension, and the attributes 0.5 dimension. Each object
in the cell can be represented with the formula as follows:
Qs FX. YT.) (1)
Where
I = identity code, corresponding to the spatio-temporal model
cell
X, Y, Z- three axis of spatial location coordinates system,
which represents the object spatial distribution, size, shape, and
topology relations with the others
T- time information
A= attributes information of objects.
In the 4.5D representation, X, Y, Z and T coordinates axis are
perpendicular with each other, and show the characteristic
continuous changes of the spatial location and time state for a
object. But, the attributes are not rigorously dimension and has
not obviously continuous development, so can be processed as
0.5D.
For the objects reconstructed from remote sensing images, the
unit of X-axis is meter, so as Y and Z. Resolution is the amount
of detail used to represent an image. Temporal resolution is
dependent on sampling frequency or the temporal scale at
which a phenomenon operates. We define the unit of T axis is
the day, because the airborne SAR system can provide us with
enough time resolution in day. And, there is the formula as
follows:
Zell ZT, i025... ©)
Where
i-order of the time interval.
The values of T are between the two threshold values, TB, the
beginning time for the object come into being, and TE, the end
time for it dieing out. Likely, the attributes values are between
AB and AE, which represent respectively the beginning and
ending attribute of the object, with a discrete value.
3.3 Model cell basic relations and accuracy analysis
Each cell in the spatio-temporal model is defined by a
succession of periods indicating different landscape states,
which are time stamped by the day that it represents. Although
it is difficult to describe directly the relations between those
cells, we can study the projection of the objects in the cell on
the axes to open out it.
Thc projection of a object on X, Y, Z-axes show the spatial
location, and the changes, including the engendering, dieing,
extending, deflating, and moving of spatial attributes over time,
will form a continuous process which is the spatial connection
relationship between two adjacent model cells when
reconstructed from two temporal high-resolution airborne SAR
images.
On the other hand, the projection of an object on T-axis has the
form of time interval. An example is a remote sensing image
time and the time when the same image is recorded could be
defined by a linear relationship. And each time interval can be
expressed with (T;, Tis) in two nearby model cells.
In addition, the relations between attributes of model cells are
more complicated, which will include the changes including
land use types in the regions, ownerships of the buildings and
so on.
As a simulation process, from static to dynamic state, the
spatio-temporal model reconstruction brings the problem of the
accuracy standard. And three types of potential errors have been
identified: temporal, spatial and attributes classification
assignment. Moreover, remote sensing images are often used to
assess change. For example, In order to detect change between
two images, A and B, subtract A from B to produce a sparse
