Com
e swaths
ted from
n remote
systems.
t spread
ilts in a
imber in
ering in
spectral
red line
> of both
r terrain
n quasi -
vods for
used for
sensors.
g image
S image
on from
created
ampling
spacing.
ed with
ions. To
ically or
ented by
1e initial
in scan
"he only
sired to
In cross
irrays is
ray and
/ can be
an ideal
y. For a
to the
Nyquist
ays the
fore the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
A
J | | EA TS
rT : ;
1 1
ET | !
' 1 :
1 1 7 b TE TA 3a
1 i Lait ! 1
1 I : |
ef oM te =
a 1 1
i | = bM 5
1 1 i ! 1 |
A 1 1 yo 15d
{ der S 1 : e
m jm 1 1 ts)
t ' p emm aud
1 ! ie! t 1
1 ' 1 i "n
/ 1 1 DS I^
- ' '
ES EE once
1 i i Dit : i
‘ ' ty i
i 1 x ta ! ;
1 1 C !
R44 i Rx Dome i-i
V ; ! PF = a 1
1 ! ! |
z i 1 | ; ' ! 1
^ je ! !
Oyis ee sn Pk =
= 1 +
i : ;
‘ ' !
| | | :
1 - 3
Ne Ny quist frequency p Myquist frequency
\ IH ^
Lt = 1126) "s z1/ A
N RF
MTF
MTF
0 0.5 1 v
pixel related spatial frequency
pixel related spatial frequency
Figure 1. Sampling and Nyquist frequency for linear and
staggerd CCD-line
1.3 Restoration
The staggered array approach improves the SNR due to a four
times lager pixel size but deteriorates the point spread
function. Therefore image restoration is necessary to realize the
interior resolution of the images. Digital image restoration is a
field of engineering that studies methods used to recover an
original scene from degraded observation. The solution is
based on estimation theory and the solution of ill-posed inverse
problems. Techniques used for image restoration are oriented
toward modelling the degradation, usually blur and noise, and
applying an inverse procedure to obtain an approximation of
the original scene.
1.4 Sampling and Resampling
Line scan data are influenced by flight movement. The result is
a variable stochastically disturbed sampling pattern on ground.
An additional randomly distributed change of sampling pattern
is due to topographic changes. To combine image data of both
staggered lines, a resampling and interpolation algorithm is
required. Image and sensor quality can be measured by the
evaluation of test structures in the image.
The organization of the paper is as follows. We start with an
overview about the ADS40 camera, image resolution and
improvement techniques and the resolution potential of
staggered lines. The third chapter describes the experiment,
data evaluation and results.
2. CCD-LINE CAMERAS AND STAGGERED CCD-
LINE ARRAYS
2.1 The Camera System ADS40
The ADS40 is based on the tree-line principle. CCD line
sensors provide a high ground resolution combined with a large
swath width.
The ADS40 has seven parallel sensor lines - three
panchromatic lines (forward, nadir, backward), three colour
lines (red, green blue) and one near infrared - in the focal
plane of a single lens system. The three panchromatic sensor
lines produce the forward, nadir and backward views along the
strip. This yields the stereo angles shown in table 1. Each
panchromatic line consists of two linear arrays, each with
12000 pixels, but staggered by 0.5 pixels as shown in figure 1.
The colour lines (red, green, blue), each 12000 pixels long. are
optically superimposed by a dichroitic beam splitter. The great
advantage of this approach is that the colour image is band
registered without significant postprocessing. The near infrared
sensor lines are 12000 pixels long and slightly offset from the
RGB triplet. The precise position of each pixel is known after
the calibration process.
Because of the linear sensor structure, the second dimension of
the image is generated by the aircraft movement and is
influenced by attitude disturbances. To correct this effect, exact
attitude and position measurements are necessary for each
image scan line being realised by an Inertial Measurement Unit
(IMU) mounted directly to the camera body.
During the flight, imagery, GPS position data, IMU data and
other house-keeping data are written to a removable disk pack.
In order to assure the desired high number of detector elements
per line, staggered arrays are used. These detectors consist of
two single [2k CCD lines positioned with a across-track shift
of half a pixel to each other.
Table 1 shows the most important parameters of the camera.
Focal length 62.7 mm
Pixel size 6.5 um
Panchromatic line 2 x 12.000 pixels
Colour lines 12.000 pixels
Field of View (across track) 64°
Sterco angles 16°, 26°, 42°
Dynamic range 12 bit
Radiometric resolution 8 bit
Ground sampling distance (3000 m 16 cm
altitude)
Swath width (3000 m altitude) 3.75 km
Read out frequency per CCD line 200 — 800 Hz
In flight storage capacity 600 GByte
Table 1. Parameters of the ADS40 camera
2.2 Image Degradation
Images are degraded by both blur and additive noise. The blur
can be described in terms of the Point Spread Function (PSF)
and is caused e.g. by the atmosphere, motion of the object or
the system, defocusing of the optics, finite size of the aperture,
the detector and so on. The measured signal in a linear model
is the convolution of the object radiance with the system PSF.
Additional noise will be superposed by the detector and
preamplifier.
The sampled signal values l'(xi,yj) (Xi = LA, yj 5j A, 1j = 0, =1,
X2.) can be obtained by a convolution of I(x,y) with the
geometrical pixel PSF
I'(x., ) = [ju (x, may, -y / (x. ya dt Ex)
H(x,y) is the space invariant point spread function, I(x.y) the
scene on detector pixel position x-iA;, y=jA,, and & the noise
of the imaging system. With abbreviation f;;-f(1^1,jA») follows
