The International Archives of the Photogramme try, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
1384
(DEM). In this step we used 50 x 50 meters resolution DEM
due to it is available freely. In order that, 3D coordinate
calculation was done using triplet images of ALOS PRISM. It
is possible to apply the calculation to other high resolution
optical sensors mounting with pushbroom scanners such as
SPOT 5 and Quickbird since ALOS PRISM sensor had same
function with them. At last, the generated DSM is evaluated by
validation points. In the registration process, 14 ground control
points and 8 validation points were introduced; resulting
X=0.97, Y=0.99 and Z=1.85 pixels in accuracy. This result was
produced by stereo matching between nadir and backward
looks. Even though the result measurement is bias with some
errors, this result could be improved by triplet image matching
and adding more validation points in future.
As of main a requirement, the study is investigated the potential
of usage of ALOS-PRISM DSM in slope failure disaster
monitoring with the area consistency of 10 square meters. As
detectable region of interest, it is possible to detect greater than
5 square meters of the area because the spatial resolution of
input image is 2.5 meters according to spatial resolution of
ALOS PRISM sensor. Finally, changed detection result is
evaluated. In the evaluation process, we created evaluation
grids. Those grids are used to measure statistic values of the
result image. Recently, this research work is applying to the
land displacement monitoring project to prevent risky landslide
disaster in the mountainous area. Future, we could be applied
triangulated irregular network (TIN) model to time series DSM
data sets together with evaluation grid statistic measurement.
2. SUDY AREA & DATA INFORMATION
2.1 Study Area
Study area is located on 4 th isle of the Japan. Geographically, it
is a small portion of Kochi Prefecture of Shikoku Island and it
had uneven elevated area. The area is one of the most landslide
risk areas of Japan by proving many geological faults. Many
landslides are happened in this area.
2.2 Data
There are two datasets are used in this study; a low resolution
DSM data and ALOS-PRISM triplet data. The PRISM data is
used to generate new DSM to measure land-form changes and
land displacement. Some characteristic of ALOS-PRISM data is
listed in the table 1.
Orbit
Sun-Synchronous Sub-Recurrent
Repeat Cycle: 46 days
Sub Cycle: 2 days
Altitude: 691.65 km (at Equator)
Inclination: 98.16°
Attitude
Determination
Accuracy
2.0 x 10‘ 4 degree (with GCP)
Position
Determination
Accuracy
lm (off-line)
Data Rate
240Mbps (via Data Relay Technology
Satellite)
120Mbps (Direct Transmission)
Number of Bands
1 (Panchromatic)
Wavelength
0.52 to 0.77 micrometers
Number of Optics
3 (Nadir; Forward; Backward)
Base-to-Height ratio
1.0 (between Forward & Backward)
Spatial Resolution
2.5m (at Nadir)
Swath Width
70km (Nadir only) / 35km (Triplet
mode)
S/N
>70
MTF
>0.2
Number of Detectors
28000 / band (Swath Width 70km)
14000 / band (Swath Width 35km)
* PRISM cannot observe areas beyond 82 degrees south and north
latitude.
Table 1: Characteristic of ALOS-PRISM data
PRISM sensor is equipped with three optical independent
systems to acquire digital surface model (DSM) with high
spatial resolution. Three temperature stabilized radiometers of
optical systems are pointing to forward, nadir and backward
views. Thus, the sensor is set to generate 5 meters elevation
data which is corresponded to the 1:25,000 scales in
topographic map together with 2.5 meters spatial resolution. To
maintain base-to-height ratio to 1, forward and backward
radiometers are inclined ±23.8 degrees outwards from the nadir.
To adjust the Earth rotation effect, each radiometer could use
electrical pointing within ±1.2 (+1.5) degrees in cross-track.
Thus, 35 km wide triplet images are acquired without yaw
steering the satellite. Lossy compression with JPEG format is
used to compress acquired data when it transmitted to ground
station (Ohta, 2006). Pointing angle +1.5 degrees in cross-track
is confirmed by PRISM first image observation (Tadono, 2006).
The scanning geometry of PRISM sensor is shown in figure (1).
46 seconds 46 seconds
Figure 1: Scan geometry and triplet image of PRISM sensor
3. TRANSFORMATION MODELS
As the prior need of DSM generation, the selected stereo
images pair must be ortho-rectified and corresponding pixels
location should be geometrically corrected. Generally, RPC
model is applied to stereo pair to generate ortho-rectified
images and to correct pixels correspondence (Dowman, 2000;
Tao, 2001). Consider producing high accurate ortho-images,
remaining errors cannot adjust by the sensor alignment and
GCPs are necessary for precise geometric correction (Fraser,
2005; Hashimoto, 2006). To produce very precise geometric
corrected one, it is need to collect very accurate ground control
points by GPS-VRS. The standard accuracy of GPS-VRS could
be acquired below the centimetre level.