DS40
togrammetry and
f Digital Surface
matching mainly
cently developed
x urban features.
ation and special
R ADS40
first commercial
eica Geosystems.
ology for sensor
r technology, and
and multispectral
scanner principle.
| plane capture
lir and backward
r CCD lines for
n provide high-
redundancy for
al image analysis.
panchromatic and
ed Position and
x ** Corporation
ory to ensure à
he goals of high
nsists of two lines.
pixel, usually one
is used.
Focal plate
uz Jd 284
Backward view
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Facal plate
A05] 1&1 =
NIR view UE view
Figure 2. Different viewing angles of ADS40 panchromatic and multi-spectral images
All the panchromatic and multi-spectral data have a high
radiometric resolution and provide a good signal-to-noise ratio.
The characteristics of the ADS40 sensor are shown in Table 1.
A more detailed discussion of the characteristics of ADS40
sensor can be found in Reulke et al. (2000), Sandau et al.
(2000) and Tempelmann et al. (2000).
The ADS40 sensor captures imagery seamlessly along the
flown strip, i.e. 10096 of the ground surface is scanned in
several strips quasi-simultaneously, which has great advantages
in both geometry and image matching procedures, and far
superior to that in typical aerial photography. This greatly
facilitates the subsequent photogrammetric processing such as
rectification, automatic point matching, block adjustment,
DEM generation and orthophoto generation.
Focal length 62.77mm
Pixel size 6.35* mero m
Field of view across flight line 64* *
Pixels per CCD line (PAN) 2 **12000
Pixels per CCD line (RGB and NIR) | 12000
| Dynamic range 14 bits
Stereo angle (forward — nadir) 28.4**
Stereo angle (nadir — backward) 14.2**
Stereo angle (nadir — NIR) 2 es
Stereo angle (RGB — nadir) 16.1%
Table 1. Sensor characteristics of ADS40
3. MATCHING STRATEGY
The ADS40 sensor is able to acquire digital data with high
geometric and radiometric resolution. All ground objects are
recorded in three panchromatic and four multi-spectral images
from different viewing angles, resulting in a redundancy in the
geometric reconstruction. This redundancy is of great
importance to the reliability of automated generation of DSMs.
Although it is possible to perform matching using more images
(panchromatic, RGB and NIR), in this study a triplet matching
using only panchromatic images is performed.
A hierarchical coarse-to-fine approach has been chosen for
triplet matching due to its advantages that it is fast and simple.
Matching is performed at pixels of extracted features mainly.
because they are the abstract of the scene; also, the processing
is fast and robust. Multi-view matching can derive a robust
approximation through the intersection of more than two image
rays. It also can increase the precision and reliability, and has
less problems caused by radiometric differences.
3.1 Image pre-processing
Due to the high dynamics of an airborne environment, the raw
images (Level 0) have to be rectified. The tight integration of
GPS, IMU and focal plate allows GPS and IMU data to be
recorded together with the ADS40 high-resolution
panchromatic and multi-spectral images during the fight in
order to facilitate ground processing. GPS/IMU data from the
Applanix system are post-processed to provide orientation data
for each image line. The camera calibration and the orientation
data are used to generate rectified ADS40 images (Level 1).
The ADS40 level 1 images are rectified onto a height plane and
the differences of scale, rotation and shear that might exist in
raw images (Level 0) are removed to a large extent.
Rectified images (Level 1) of three panchromatic channels are
further processed by using a Wallis filter in order to obtain a
better contrast enhancement.
3.2 Generation of image pyramids
A fast matching approach is very important for practical
applications, especially when a multi-view matching is
performed, in which a large amount of data has to be processed.
For ADS40 high-resolution images, one strip corresponds to
three panchromatic channels (forward, nadir and backward).
Since each image could be kilometres long, which corresponds
to more than 1 Gigabyte, the processing time should be
considered and the memory should be efficiently managed. In
order to manage and process this large amount of data
efficiently, a coarse-to-fine matching strategy may be a good
choice. In a hierarchical coarse-to-fine architecture, images are
represented in a variety of resolutions, leading to an image
pyramid. Results achieved on one resolution are considered as
approximations for the next finer level. Thus, the search range
in each level can be restricted within a very small area and the
matching process is fast. The upper levels of the pyramids are
ideal to get an overview of the image scene. The details can be
found down the pyramid at higher resolution. The coarse-to-
fine strategy also reduces the necessity for initial values for the
points to be matched.
