Vol. XXXVIII, Part 7B
In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
105
:tober 31, 2009; 10:00
5 solar zenith, 137.6 0
s the lOlOnm band, C
TOnm) interpreted as
effect have applicable
and processing that is
Dual images, acquired
ed campaigns. Three
illow: 1) enhancing
kV, and 3) performing a
ie-shadowing process
ge's statistics to gain
' to correct the shadow
data sets acquired on
ocess is missing in the
ing algorithm consists
teps: an atmospheric
isking, five additional
iriance matrixes and
task, and final step is a
to the pixels in the
ping shadow areas in
An interpretation of
ow areas with highly
l' for each pixel in the
e provides an external
i of ATCOR-4, allows
fication. The proposed
y tested on six scenes
ltage of the presented
operator, and it is fast
on the boresight ratio
Figure 3. Shadow masks of - A is an AISA-Dual image
(December 18, 2009; 11:30 GMT; 15km visibility; midlatitude
winter model; 60 0 solar zenith; 205.8 0 azimuth angle), B is a
boresight shadow mask, C is a boresight shadow estimation
(yellow 0.1-20%, cyan 20-50%, magenta 50-80% and maroon
80-100%), D is an ATCOR-4 shadow mask
3.2 Stereo 3-D map
The simultaneous across-track stereo-data acquisition gives a
strong advantage in terms of radiometric variations. Since an
error of ±3 pixel along-track and ±1 pixels across-track for the
parallax measurements in the automated matching process has
been achieved with these different datasets (along-track and
across-track), the potential accuracy for the across-track stereo-
derived local DEM from AISA-Dual could be on the order of 3
m (1.5x1.5 pixels) . The main objectives of this application are
to generate and evaluate pixel based local DEMs from the
boresight VNIR and SWIR images. The 3-D stereo intersection
is performed using a computed geometric model to convert the
pixel coordinates in both images determined in the image
matching of the stereo pair to 3-D data. The non earth
coordinates are determined for the measured point with a least
square intersection process based on the geometric model
equations and parameters (Toutin, 1995). The result is an
irregular grid in the map projection system, which is
transformed to a raw regular DEM.
Figure 4. DEM, A is the AISA-Dual image (October 122, 2006;
08:55 GMT; midlalitude summer model; 46.4 0 solar zenith;
164.6 0 azimuth angle), B is the 3-D stereo model, C is the
boresight calculated band, D is the raw DEM (cloud masked-
black border) modeled with Inverse Distance Weighting (IDW)
algorithm (0m 30m), E is the AISA-Dual in 3-D view
(x axis 1500m, y axis 500m, z axis 30m)
3.3 Unmixing and anomaly detection
Most of the pixels collected by HRS airborne sensors contain
mixed spectra from the reflected surface radiation of various
materials in the sub-pixels. As a result, mixed pixels may exist
when the spatial resolution of the sensor is not sufficient to
separate different pure signature classes. The resulting spectral
measurement is a composite of the individual pure spectra
weighted by a set of scalar endmember-abundance fractions
(Adams et al., 1986). Under such circumstances, target
detection must be carried out at sub-pixel level. An anomaly
detector enables to detect targets whose signatures are spectrally
distinct from their surroundings with no a priori knowledge. In
general, such anomalous targets are relatively small compared
to the image background and only occur in the image with low
probabilities. Two approaches are of particular interest. One
was developed by Reed and Yu (Reed and Yu, 1990; Yu et al.,
1993; Yu et al., 1997) and is referred to as the RX detector
(RXD). The RX detects spatial/spectral anomalies using the
sample covariance matrix to detect "interesting target" pixels
which occur with low probabilities in the data (i.e., the size of
target samples is small) compared to Gaussian distribution of
the background. Second is a SVDD, (support vector approach)
that is a non-parametric method with several benefits, including
non-Gaussian modeling basis that can model arbitrarily shaped
and multi-model distributions, scarcity and high generalization
ability (Baneijee et al., 2006).
We suggest using boresight calculated band as spectral/spatial
anomaly detector. Since an anomaly defined as a small target
with distinct spectrum, an accurate pixel-by-pixel classification
of boresight values may accentuate the targets in question.