The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part Bl. Beijing 2008
1119
GPS observations and post processing of the data in differential
mode.
Figure 1. Methodology flowchart for stereo data analysis,
validation and extraction of terrain parameters
Post processing of GPS observations was carried out using
Static Kinematic Interface (SKI Pro V 2.1) software. GPS data
of six International GPS services for Geodynamics (IGS)
stations, having a distance of 3000 km or less from the
reference station, was downloaded in RINEX format along with
precise ephemeris data. The reference station was established
by network adjustment with IGS stations taking 3 days data. A
sub-reference station was established in the study area with
respect to reference station by taking observations for 24 hrs.
The GCPs were established by base line processing with respect
to sub-reference station. The common observation period was
kept 1 hour between the sub-reference and GCPs. In all 19
GCPs were established. All the coordinates were in WGS-84
datum. The coordinates of GCP’s derived from differential GPS
measurements were validated using vector closure method at
three non-collinear stations, reference station being one among
them. The closing error of the triangle was computed by (V
(Edx 2 + Edy 2 + Edz 2) / ES), where Edx, Edy, Edz represent the
sum of component vectors in x, y and z direction respectively
(the misclosures in three directions) and ES represents the sum
of the sides of the triangle. The closing error is better than 1
ppm.
Effect of Number of GCPs and Polynomial Order used in
RPC Refinement, for Generating the DEM
GCPs are the inevitable components for generating an accurate
DEM and Ortho images. Ortho rectification can be performed
using RPCs provided along with data, with out any ground
control information. These products will lack in the accuracy
part. Leica Photogrammetric Suite v. 9.0 (LPS) was used for
photogrammetric analysis in this study. Out of 19 GCPs, used
in this study, 10 were used as planimetric and vertical control
points, 7 were used as planimetric and vertical check points and
2 were used as vertical check points. More than 250 tie points
were automatically generated. GCPs were used to refine the
RPCs for the generation of DEM and the accuracy was assessed
at check points.
Initially, no GCPs were used for refining the RPCs and
planimetric and elevation accuracy at 9 check points was
computed by triangulating. Number of distributed GCPs was
varied by 1, 5, 8 and 10 and also the order of polynomial
refinement. With each group of GCPs triangulation was
performed for different polynomial order from 0 to 2 depending
on the number of GCPs. For 0 and 1 GCPs the polynomial order
0 was taken. Accuracy at GCP and Check points were
computed. The triangulation exercise was carried out by taking
polynomial order 1 and changing number of GCPs as 5, 8, and
10. This was extended for polynomial order 2 for GCPs 8 and
10. The results are tabulated in Table 2.
No.
of
GCPs
Poly.
Orde
r
RMSE at Control
Points (10 Full)
(m)
RMSE at Check
Points
(7 Full + 2 Vertical)
(m)
Lon.
Lat.
Eie.
Lon.
Lat.
Eie.
0
0
22.6
98.4
22.9
1
0
9.8
4.4
4.5
5
1
3.7
4.0
0.4
4.9
3.9
3.5
8
1
3.7
2.7
1.3
6.6
4.1
2.8
8
2
0.4
0.7
0.4
3.6
6.3
3.2
10
1
5.4
3.2
1.3
5.0
3.6
2.9
10
2
0.4
0.7
0.4
4.3
3.5
3.1
Table 2. Effect of Number of GCPs and Polynomial order for
refinement of RPCs for DEM generation.
DEM and Ortho Image Generation
A combination of the order and number of GCPs having
minimum RMS Error was selected for generating final DEM.
Figure 2. DEM for the study area
DEMs were generated at 2.5m, 5m and 10m pixel resolution for
a subset. Subset was selected to reduce the number of irregular
points for interpolation. Around 40 Lakh points were generated
for the entire study area at 2.5 m resolution. Figure 2 & 3 shows
the DEM and Ortho image generated for the study area.
