Istanbul 2004
‘ule
nent)
ice for image
hcentrate on
ments and
nented in an
cel size of 20
verage of 9,2
r wide angle
ig height
m
, 120
2,200
image scale
conventional
multancously
directly in a
n for images
'craft's flying
ely by taking
No. of
spectral
bands
11
100 -
rborne sensor
'€ area on the
| bands. The
. whereas the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol XXXV. Part B2. Istanbul 2004
number of spectral bands is of major importance in image
classification.
2.4 Satellite sensor systems
The commercial high resolution satellites systems are the third
group of sensor systems which might be useful for LPIS. The
most important ones are characterized in the following table:
| Satellite Ground | Pan MS Rev.
swath res. res. time
[km] [m] [m] | [days]
IKONOS ; ;
(Space Imaging, 2002) H ! A ns
QuickBird ; ar
(DigitalGlobe, 2003) loc 4091.1 2:34 lid 3
SPOT 5
2 =
(Spotimage - b) by 2.3 10 Peu
EROS AI i
3.55 | 18 1. — 3
(ImageSat, 2003) 133 Ls 2-3
Table 3. Satellite sensor systems
IKONOS and QuickBird would fulfill the requirement of a 1-
meter resolution (pan sharpened) and therefore have been taken
into consideration for LPIS. SPOT 5 with a resolution of 2.5
meter in "Supermode" (Spotimage-a, 2003) does not fulfill the
requirements, but would have the advantage of a wide ground
coverage. Also not qualified is the 1.8 m resolution imagery of
EROS Al.
3. IMAGE RECTIFICATION
As already mentioned, images used in LPIS should be ortho-
rectified images. The two main factors influencing the accuracy
of the rectified images, the
e Height information and the
* Leaning effect
will be discussed in the following. The influence of image
orientation is almost similar for all image recording systems,
thus a discussion of the orientation impact is omitted.
3.1 Height Information
The influence of height on the accuracy in orthophotos is basic
knowledge covered in all textbooks of Photogrammetry. Based
on the assumption that all, airborne or spaceborne, images are
differentially rectified using digital elevation models the impact
of a height error on a location in a orthoimage can be estimated
by
AZ
AR = (1) (Kraus 1989)
c/p' + tan à * cos B
where AR = location error
AZ = height error
¢ = focal length
p' = distance from the image center to location of the
considered object location
a = terrain slope
B = angle between the straight line beginning in the
image center to the considered object location
'A
-
The following table (Table 4) shows position errors, calculated
based on:
* 10-meter height error as an example
« ß=90°
e image corners / edges considered
e a ground slope of u =30° was chosen. This is the
maximum slope agricultural machines can work on.
System u Position error
[degree] [m]
Wide Angle Camera 0 9,15
Wide Angle Camera - 30 19,40
Normal Angle Camera 0 4,59
Normal Angle Camera - 30 6,25
IKONOS nadir 0 0,08
IKONOS nadir - 30 0,08
IKONOS 30? pointing 0 5,14
IKONOS 30? pointing - 30 7.31
SPOT nadir 0 0,37
SPOT nadir - 30 0:37
SPOT 27° pointing 0 5.13
SPOT 27° pointing - 30 7,30
HRSC - AX 0 2,62
HRSC - AX - 30 3.08
ADS 40 0 6,24
ADS 40 - 30 9,75
Table 4. Position errors, caused by height errors
(for more details please refer to Oesterle 2003)
Obviously the position error gets smaller if a system with a
large focal length is used. Therefore it can be concluded, that
the use of systems with a good relation of ground coverage to
flving height will benefit the production of accurate
orthophotos. Highly accurate height data almost climinate this
type of error.
3.2 Leaning Effect
This effect is comparable to the position error, but calculated for
flat terrain (alpha =0). The leaning effect is caused by depicting
a 3D object on the 2D image. It can be observed at objects,
protruding the earth's surface such as buildings and forest
borders. The leaning effect can be avoided by producing True
Orthophotos. However, this process requires measurements of
the 3D objects. Furthermore it requires the use of overlapping
image areas, in order to get information about hidden regions.
The production of True Orthophotos covering the usually quite
big area of a state is less recommended, as the measurement of
3D object information is time consuming and therefore
expensive. Airborne Laserscanning may help to reduce the costs
of providing the data for True Orthophoto generation.
Figure. 3. Building shown with leaning effect, compared to
"True" building representation