OPERATIONAL DETERMINATION OF TIE POINTS AND BUNDLE ADJUSTMENT OF 
HRSC IMAGES OF THE MARS EXPRESS MISSION 
R. Schmidt 1 *, M. Spiegel, C. Heipke 3 , A. Dumke b , G. Neukum b and the HRSC Co-Investigator Team 
institute of Photogrammetry and Geoinformation, Leibniz Universität Hannover, Nienburger Str. 1, D-30167, 
Hannover, Germany - schmidt@ipi.uni-hannover.de 
b Planetary Sciences and Remote Sensing, Freie Universität Berlin, Malteserstr. 74-100, D-12249 Berlin, Germany - 
michael.spiegel@mytum.de 
Commission IV, WG IV/7 
KEY WORDS: Extra-terrestrial, Adjustment, Matching, Orientation, Processing, Bundle, Three-Line, Planetary 
ABSTRACT: 
The pushbroom scanner HRSC (High Resolution Stereo Camera) onboard the European Mars Express mission is orbiting the planet 
Mars since January 2004 and delivers stereoscopic imagery with five panchromatic and four colour channels. This paper describes 
the process of improving the exterior orientation of the HRSC which is needed in order to derive high quality products such as high 
resolution Digital Terrain Models (DTMs) and ortho image mosaics from the data. The systematic photogrammetric processing for 
improving the exterior orientation is divided into two steps: First, a large number of tie points have to be automatically determined 
in the images using digital image matching. Second, a bundle adjustment is carried out using the globally available MOLA (Mars 
Orbiter Laser Altimeter) DTM as control information. The developed approaches of the matching and the bundle adjustment will be 
described in this paper. After that, the performance of both methods will be evaluated with respect to point accuracy and 
consistency with the MOLA DTM on the basis of 45 test data sets. 
1. INTRODUCTION 
Since January 2004 the HRSC (High Resolution Stereo Camera) 
onboard Mars Express is imaging the planet Mars 
stereoscopically, in colour and at high resolution and has 
returned more than 2300 image strips to earth. The pushbroom 
scanner, equipped with 9 CCD line detectors with 5176 pixels 
each mounted in parallel on the focal plane, is able to obtain 
panchromatic stereo data at three or five angles and colour data 
simultaneously in one image strip (Neukum et al. 2004). With 
an average flying height of about 270 km at pericenter a 
resolution of up to 12 m per pixel is achieved. The three- 
dimensional position of the orbiter with respect to the body- 
fixed coordinate system of Mars is permanently observed by 
Doppler measurements of the FDT (Flight Dynamics Team) of 
ESOC (European Space Operations Centre). The pose of the 
camera is adjusted with star trackers to the viewing direction 
from mission planning. The two elements position and pose 
form the exterior orientation of the HRSC with six values for 
each image line. An absolute accuracy of about 200 m is 
achieved in practice, but along flight direction the deviation can 
amount up to a few hundred meters. These values are not 
accurate enough for precise photogrammetric point 
determination. Therefore, a bundle adjustment is carried out to 
improve the relative accuracy of the ray intersections and the 
absolute position of the spacecraft. For absolute positioning the 
globally available MOLA (Mars Orbiter Laser Altimeter) DTM 
(Neumann et al. 2003) is used, which features high global 
accuracy and currently marks the best available reference 
system on Mars. With the improved exterior orientation high 
quality products such as high resolution DTMs or ortho image 
mosaics can be produced. 
The improvement of the exterior orientation of the orbiter is 
divided into two separate steps: At first, a large number of tie 
points have to be determined automatically which will be used 
in the second step as input in the bundle adjustment. The 
automatic extraction of tie points is carried out at Institute of 
Photogrammetry and Geoinformation (IPI) of Leibniz 
Universität Hannover and the result is transferred to 
Photogrammetry and Remote Sensing (FPF) group of 
Technische Universität München respectively the Planetary 
Sciences and Remote Sensing group of Freie Universität Berlin. 
Chapter 2 of this paper describes the process of automatically 
determining tie points in HRSC images. In chapter 3 the 
method of the bundle adjustment is explained. In chapter 4 the 
results of the single image strip processing are presented 
exemplified with 45 test orbits. The results of processing 
photogrammetric blocks are described in an accompanying 
paper (Dumke et al. 2008). Chapter 5 presents a short statistic 
of the results of the operational standard processing. The last 
chapter summarises the results and the conclusions are drawn. 
2. DETERMINATION OF TIE POINTS 
In this chapter the automatic determination of tie points is 
described. The general processing chain is displayed in Figure 
1. The starting point are the radiometrically corrected HRSC 
images (called “level-2 images”). Because HRSC images often 
have a low signal-to-noise ratio an optional low pass filtering 
can be applied (cf. Gwinner et al. 2005). This step increases the 
robustness of the matching and the number of tie points. 
Extensive tests (Schmidt 2008) have shown that denoising 
algorithms like Anisotropic Diffusion (Perona & Malik 1990) or 
Corresponding author.