4.1.8. Multi-Media Photogrammetry
Today, bundle triangulation programs are usually
based on the functional model of collinearity. The
deviations from ideal central projection (e.g. lens
distortion, film unflatness and deformations etc.)
are modelled by additional image error describing
parameters. Such a functional model is usually in-
appropriate for applications where light-refracting
surfaces are located between the object and the pro-
jection centers, e.g.:
» under water photogrammetry,
- mapping in shallow water,
+ survey of sheltered objects,
correction for errors generated by reseau plates.
In those cases the functional model must be ex-
tended for considering light refracting surfaces
(Kotowski, 1987, 1988).
Inspired by pratical requirements CAP includes a
sophisticated mathematical model for mult-media
photogrammetry.
4.1.9. User Interface
CAP is a batch program. Once the program is
started the user doesn't need to interact any more
with the adjustment program. But he is constantly
informed on the computers screen about the current
status of the program.
4.1.10.Performance of CAP
To give an impression of the over all performance
of the algorithms implemented in CAP the table 1
below shows some results of computation times of
projects of various size. The over all computation
times include finding the adjacency structure of the
block, solving the least squares adjustment by New-
ton Gauss method, computation of the inverse of
the normal equation matrix, computation of all sta-
tistical values and saving all results to output files.
Example: A B C D E
# Obj. Points 198 | 58 | 502 | 216 | 84
# Images 13 12 17 24 38
# Geod. Observ. 7 17 44 23 S
# Camera Params. 5 5 5 S 5
# Unknowns 684 | 257 |1,613| 803 | 492
Total Redundancy | 1,497 | 744 | 1,963 | 2,050 | 1,015
# Iterations 6 5 4 4 5
Time [sec] 220 | 108 | 284 | 354 | 436
Table 1: Performance of CAP
All numerical computations were carried out on a
Compaq 386/20e which was equipped with a 80387
math coprocessor.
5. ANALYSIS OF ADJUSTMENT RESULTS
Besides the standard ASCII file output created by
CAP there exists a special interface to exchange
data with the analysis program GRAV.
16
GRAV is a highly interactive program. When the
program is started it will automatically retrieve the
data of the last bundle adjustment. The first view
provides the user the object coordinates and loca-
tions of projection centers in the X-Y plane of the
object coordinate system. The example used
throughout this chapter shows an object in the cen-
ter of the ficld of view and the projection center
locations around the whole object.
XY
Number of Links .
Connections
A; Activele ...
D: Deactvate …
V: en
F2: Zoom
F3: Reset Zoom
Figure 1: Object and projection centers in X-Y
plane.
The user can view the object from any coordinate
plane. For better overview the point and image
numbers can be displayed selectively. To distin-
guish between different types of object points and
projection centers the user can define color, shape
and size of any type which is displayed. On a color
monitor this makes the analysis of the results very
convenient.
Vez
Ni
c ->
xz
Y-2
Nemes ...
Number of Links .
Connections
A: Activele ...
D: Deaciivete ...
v:
F2: Zoom
F3: Reset Zoom
Figure 2: Display of object in Y-Z plane. Image
numbers activated and some rays from images to
object points.
The upper left corner of the display region displays
some status informations. This view shows that the
images were taken from two height levels.
A very useful feature for comparison of points of
interest is the so called split frame mode. In this
mode the display region is subdivided into four
small windows. Each of the windows can display
different i
ment.
XY
NI
c
Figure 3:
At the sar
locations
coordinat
image nu
that imag
The displ
on the lar
rmsX: 8.338e-00:
tms: $.5356-00:
rl.
\ —
»
x -30.
y: -30.
Figure 4:
big frame
In the up
rms value
color, sh
respond t
uals are
This is v
uals. Blu
image are
tionally 1
circle.
To analy:
can be zo
The few
monstrate
for the an
