1363 
GEOMETRIC VALIDATION OF CARTOSAT-1 IMAGERY 
S. Kocaman*, K. Wolff, A. Gruen, E. Baltsavias 
Institute of Geodesy and Photogrammetry, ETH Zurich 
Wolfgang-Pauli-Str. 15, CH-8093 Zurich, Switzerland - (skocaman, wolff, agruen, manos@geod.baug.ethz.ch) 
Commission I, SS-11 
KEY WORDS: Satellite Remote Sensing, Digital Surface Models (DSM), Image Matching, Adjustment, Topographic Mapping, 
Image Processing, Accuracy Analysis 
ABSTRACT: 
The Institute of Geodesy and Photogrammetry (IGP) is participating in the Cartosat-1 evaluation program, a common initiative of 
ISRO (India) and ISPRS. Within this program, various test sites with reference data have been established and Cartosat-1 images 
have been acquired over these sites. We have reported already about the Rome and Mausanne datasets in a former publication. Here, 
we report about our investigations at the Catalonia test site of ISRO. In addition, we evaluate the Cartosat-1 images acquired over the 
test site in Sakurajima, Japan. First, we report on image preprocessing for their improvement. Then, we report on RPC refinement 
and the 3D point positioning accuracy that can be achieved with the Rational Polynomial Coefficients (RPCs) sensor model. The 
orientation results include various options regarding sensor model and number and distribution of GCPs. The best results led to 
planimetric and height accuracies (RMSE) of about half a ground pixel size (2.5m) in both planimetry and height. Finally, we 
present the results of automatic DSM generation using our own SAT-PP program package. The results were checked both visually 
and were compared using the provided reference data (Catalonia DTM and Sakurajima DSM). While for the Catalonia test site the 
achieved height accuracy is about 3 m (sigma), confirming previous results we have reported in other test sites, for Sakurajima the 
results are significantly worse mainly due to the terrain relief. 
1. INTRODUCTION 
High spatial resolution optical satellite sensors have been 
subject of scientific investigations and evaluations since 2000. 
Some of them provide a ground sampling distance (GSD) of lm 
or less but each image covers a small area, their price is high, 
and stereo coverage is often rare. Thus, the last years some 
sensors have been launched with a GSD of 2.5 -5 m, covering 
a much larger area per image, having much lower image price 
and tailored to acquisition of stereo images and derivation of 
DSMs by using 2- or 3-Linear Array CCD sensors. Typical 
examples include Spot-5 HRG, ALOS/PRISM and Cartosat-1. 
Such systems are suitable also for derivation of global DSMs, if 
the absolute geolocation accuracy of the images is good enough 
(e.g. SPOT-5, ALOS/PRISM). 
A description of the Cartosat-1 mission is given in 
Krishnaswamy and Kalyanaraman (2002), Knshnaswamy 
(2002), Cartosat-1 Handbook (2006), EOPortal (2007) and 
Cartosat (2008). Here, we remind some of the parameters, 
which are relevant for the discussion below. Cartosat-1 has a 
fore (F) and aft (A) panchromatic camera for along-track stereo, 
with a tilt in flight direction of +26° and -5°, respectively. Each 
sensor is comprised of 12,000 pixel CCDs with 7 microns pixel 
spacing. Each sensor uses two staggered lines, which take 
images with a time difference of about 1.7 ms (the distance 
between the two lines in flight direction is 35 microns). The aim 
of the staggering is not the increase of spatial resolution (only 
odd and even pixels are read-out) but a faster read-out, 
permitting the integration time needed in order to achieve a 
pixel footprint of 2.2-2.5 m in flight direction. The CCDs, 
identical to the one of Cartosat-2, are not of TDI (Time Delay 
and Integration) technology and the integration time is 0.366 ms, 
so rather short compared to the usual target value of 1-2 ms. 
The nominal base to height (B/H) ratio is 0.62. Data are 
quantized with 10-bit and compressed by a factor of 3.22, with 
little image quality loss. The nominal GSD is 2.2-2.5m, and the 
typical image size is 12,000 x 12,000 pixels. The image scale is 
about 1:312,000. The geolocation accuracy of the images, as 
given by ISRO, (without GCPs, 3-sigma) is 250 m (design) and 
150 m (achieved). The satellite has a yaw steering, in order to 
compensate the Earth rotation effect or to acquire a wider mono 
strip. A pitch bias (e.g. 5 deg, -21 deg, -10.5 deg) is also 
possible to acquire occluded areas in case of large slopes along 
track. A roll of the satellite body allows across-track pointing 
up to 23 degrees, increasing thus the revisit frequency. 
Dynamic changes, especially the continuous yaw steering and 
pitch bias (if applied during imaging) may affect both 
geometric stability and radiometric quality (image smearing). 
A set of algorithms for processing of high resolution satellite 
imagery (HRSI) has been developed at the Institute of Geodesy 
and Photogrammetry (IGP), ETH Zurich and realized in a 
software suite called SAT-PP (Satellite Image Precision 
Processing). The SAT-PP features mainly include: GCP 
measurements, image georeferencing with RPC approach and 
various other sensor models, DSM generation with advanced 
multi-image geometrically constrained matching for linear array 
and frame sensors, ortho-image generation, and feature 
extraction. The software has been used for processing of a 
number of high resolution satellite sensors, such as IKONOS, 
QuickBird, ALOS/PRISM, and SPOT-5 HRS/HRG. Detailed 
information on the SAT-PP features can be found in Gruen et al. 
(2005). 
Corresponding author.