International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
positional sensors. The main reason for the aerial triangulation
process is to reduce the cost of ground control point surveying
by replacing it with computed coordinates from image
observations.
i5 TA Points Data
pum Tem Fe =_ S ~~" Read Image Data
€
Calibrate Image is
Sensor =
# ^.
# { e n E
Sd Bundle Adjustment Least Squares
(sl Adjustment
se
A
Imaging Sensor Pel
Spars = Ee ;
\ S. Control Imaging
, 3 Sensor s
1 x
^ Cho ers
-——-. Transaits hage CBia "A 5
1 if i
\ Log mage Data maging
\ Sensor
i e
\ J
} T
M /
E Trad su its Shutter Pulse
Monitor Flight J
oh
Record Shutter
Pulse
France is 1 PPS
i
a s. Transmit Fange Data D
CHO C m A.
Compute Present Log GPS Data GPS
Location Receier
Figure 1. Use Case Diagram of the Image Acquisition
Subsystem.
During the aerial triangulation process, the operator identifies
the surveyed ground control point in the images in which they
appear and measures the image coordinates of these points.
These measurements and the surveyed coordinates are then
fed into an aerial triangulation program to compute the
coordinates of the perspective points as well as the attitudes of
the observed and other connected images at their instants of
exposure. These (i.e. the image coordinates, the perspective
centre coordinates and image attitudes) are then used to
compute ground coordinates of any points on an image.
In the triangulation process of an automated mapping system,
the perspective point coordinates and the attitudes, are not
computed indirectly, but are observed directly using the
positional sensors. In theory, the triangulation process would
not involve any ground control point surveying.
Aside from the fact that the perspective coordinates and the
attitudes are directly observed, the setting up of the
observation equations and its adjustment is very similar to the
traditional aerial triangulation task.
2.4 Object extraction subsystem
Object extraction refers to the interpretation and recording of
objects from images. Automation of such tasks is still a
challenge to many researchers in both the photogrammetric
field and the computer vision field. Object recognition
techniques are used by the photogrammetrist to capture the
semantic information at a certain location and populate the
GIS database, which is identical to the task of the human
stereoplotter operator. The automation effort in this field
involves research into image segmentation, feature extraction
from images and grouping extracted features such as points
and edges. Detection and interpretation of simple features such
as road centrelines has been successful, but other spatial
objects, buildings in particular, are still being researched
(Roux et al, 1994), (Gruen et al., 1996), (Boichis et al., 1998)
(Haala et al., 1998).
( 3
^ X
_ Point Pos itioning with =
en Code Ranges xs:
— ^
\ 77? Least Squares
s Adjustment
Kinematic Phase a fi
~~ Processing c |
Static Phase /
M
re S zs. ; he 2
an i eadiMU Data — > Processing {Interpolate Position
Positioning Sensor | / S me 7
/ . S 1 1
Specialist = MCN Es DOS
NS n = i
| wT Tes a Hl
| (7 7 Red GPS Daa ( E
1 A S ertet S
Read pais File — SmosfiDala rue Data
| Us m
| 2 S A i A
\ | ros / 7
1 ! P MA 7 /
i | wii j
1 Read Event Data,” =
j A.
i Er j
Lr Me chani e IMU Data
m #
Zero Velocity Update /
¢
i
^
Process Inu Data IU Aart
Figure 2. Use Case Diagram of the Positioning Subsystem
2.5 Visualization subsystem
Schroeder (1998) defines visualisation as "... the process of
exploring, transforming, and viewing data as images (or other
sensory forms) to gain understanding and insight into the data
". The concept of automation and the involvement of
computer processing is inherent in this definition of
visualisation. This definition resembles closely the activity of
a cartographer, except that a cartographer deals mainly with
geospatial data.
The end product of a visualisation process in a mapping
system might be a three dimensional perspective view of a
landscape processed from the digital terrain model and a
scanned aerial image of that area. The images of the
mountains could also be marked with contour lines, streets
could be labeled with street names and commercial buildings
could be coloured red. In other words, the traditional
cartographic process of symbolising information on paper
maps is incorporated into the visualisation process as
symbolising information on three dimensional image views.
The bundle adjustment use case of the triangulation subsystem
was implemented as shown in Figure 3. The figure shows text
files opened in three child windows each showing the
attributes of the images, control points and the imaging sensor
respectively.
pa
cajliiname = st1
Ra 11} sensor = olympus
De] 12|InitialPersCenter 7 136.0 179.0 6
At 12 InitialRotation = 0.0 0.0 0.0
EN20LP
|^ J20i101 215.870968 418.935484
210102 212.967742 520.870968
21,103 209.741935 680.000000
1 643.774194 399.225006
646.354039 506.709677
649.258065 674.838710
1047.096774 32.806452
1056.129032 491.838710
23 1065.806452 661.612903
Ready
Figure 3. Implementation of the bundle adjustment use case.
3
Interi
Grapl
imprc
engin
bundl
contre
instea
x
New
DCpDOonDogonon«
i i
UE
i | 1000000000» -
|
+. !
In this
utilizes
based
Netwo
offers |
of the
exist «
constru
(Booct
describ
In this
and 3D
operati
polygo
class di
Ll
2D-
3DTINStrc
MÀ
Figure .
The x:
3DTIN(
data be
modelli;
the foll
OO TIN