The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
map contains terrain, forest, vegetation, roads, lakes, buildings,
and geographic names. Synthesizing the concepts associated
with photogrammetry, remote sensing, GIS, and 3D
visualization introduces a new paradigm for the future of digital
mapping. Correctly combining height information with existing
2D maps has a great potential for a fast, accurate and highly
automated generation of maps [2].
This paper describes a methodology to update 1:25000
topographical map using Cartosat-1 and Cartosat-2 images in an
integrated approach. The other objectives are to test the
capability of feature extraction in stereo mode from the
Cartosat-1 stereo data and comparison between features directly
traced from Cartosat-2 ortho-image. This approach include data
pre-processing, modeling of space imagery, DEM and Ortho
image generation with or without GCPs, accuracy improvement
of DEM and Ortho-image, Image to Image registration, 2D and
3D feature extraction and mapping in a standard cartographic
environment. This study also provides a guideline for the 2D or
3D extraction for different feature types.
2.0 STUDY AREA
Covering an area of 7.5’X7.5’ of the test site is bound between
geo-coordinates 17° 22’30” N to 17° 30’ N and 78° 22’30” E to
78° 30’ E. The study area forms part of Hyderabad City and the
general elevation of the area ranges from 400 m to 630 m above
mean sea level.
3.0 INPUT DATASETS
Cartosat-1 stereo images and Cartosat-2 datasets have been
acquired for the study area. Both the satellite datasets were
accompanied by respective RPC files. The verification was
made for the datasets after receiving the datasets. There were 20
Ground Control Points used for this project for modeling and
accuracy evaluation.
4.0 METHODOLOGY
A systematic framework of the algorithm used to generate
digital maps is presented in figure-1. The analysis of
methodology can be grouped into three stages. The first stage
incorporates the Cartosat-1 processing for DEM generation and
3D feature extraction. The second stage related to Cartosat-2
processing and feature extraction in 2D. The final stage used a
set of overlay decision rules to inject a knowledge-based feature
comparison based on suitable features for mapping from the
two satellites.
A project for DEM generation and stereo processing is created
in Lieca Photogrammetry Suite (LPS) S/W and the Cartosat-1
stereo images were imported. The interior orientation has been
computed for both fore and aft camera images (BandA &
BandF) using RPC. 15 GCPs have been identified on both the
images for Exterior orientation computation. Image matching
has been performed between the fore and aft images. A total
number of 6326 match points called as Tie points have been
generated at more than 0.80 correlation coefficients.
Triangulation has been performed with RPC at 2 nd order
polynomial refinement. The regular DEM has been generated at
10 m grid interval. Finally Ortho-image is generated and
features have been extracted in 2D mode by on screen
digitization. The same model has also been used for 3D feature
extraction. All the man made structures (buildings, fence etc.)
and Trees have been captured in 3D environment in stereo
mode. The accuracy achieved for DEM and orthoimage is better
than 5 m.
The Cartosat-2 image has been registered with Cartosat-1 ortho
image using the Automatic Point Measurement (APM)
Software tool [3]. APM uses image matching technology to
automatically recognize and measure the corresponding image
points between the two images. APM deliver the coordinates of
evenly distributed corresponding points between an input image
and a reference image. The search and cross-correlation
window sizes for matching were set to 17 and 11 respectively.
The window size for least square matching for this project has
been taken as 10 pixels. The limit for cross correlation
coefficient was 0.80. A total of 500 well distributed points
between the images have been generated at an RMS Of 0.42
pixels with a standard deviation of 0.25. An image to image
affine transformation was performed to register the Cartosat-2
image with respect to Cartosat-1 image. Final resampled
Cartosat-2 image has also been visually compared with
Cartosat-1 orthoimage. List of features captured in 2D and 3D
mode from Cartosat-1 and Cartosat-2 ortho-images is given in
table-1.
Figure 1 : Mapping schematic work-flow frame work
5.0 RESULTS AND CONCLUSIONS
A comparative study has been made for the features extracted in
2D from Cartosat-1 and Cartosat-2 ortho-images and it has been
observed that feature extracted from Cartosat-2 is giving better
result with respect to features shape and size. This is obvious
due to the higher resolution of Cartosat-2 with respect to
Cartosat-1. The resolution difference has been shown in figure
1(a) and 1(b). It has been observed that capturing features in the
city like Hyderabad using Cartosat-1 in 2D is very difficult.
One can only capture the building blocks from Cartosat-1
[figure-2] which depends on the density of buildings. But
buildings can be captured in stereo mode in 3D [figure-3] with
Cartosat-1 data, which nearly match with those derived from
Cartosat-2 in 2D mode.