1049 
3D CRATER DATABASE PRODUCTION ON MARS BY AUTOMATED CRATER 
DETECTION AND DATA FUSION 
J. I. Simpson a-b *, J.R Kim b , J-P. Muller b 
a QINETIQ Sundridge, Darenth House, 84 Main Road, Nr. Sevenoaks, Kent, TNI4 6ER, UK 
b Mullard Space Science Laboratory - Holmbury St. Mary - Dorking - Surrey - RH5 6NT, UK 
jsimpson@qinetiq.com, jkim@mssl.ucl.ac.uk, jpm@mssl.ucl.ac.uk 
WG VI/7 
KEY WORDS: Planetary mapping, Geology, Automation, Imagery, GIS, DEM/DTM, Fusion, Algorithms 
ABSTRACT: 
Impact crater databases are a key resource for planetary geologists. Uses include studies of relative and absolute surface (age) 
chronologies, erosional processes, hydrological evolution and climate history. This paper describes the creation of 2D and 3D Mars 
crater databases from high resolution HRSC stereo imagery using data fusion and automated crater detection methods. A semi- 
automated process has been developed which incorporates a software GIS tool to facilitate the statistical assessment of detection 
rates and quality as well as provide 100% coverage with minimal human interaction. A specialized stereo matching system for 
impact craters using high quality HRSC stereo imagery with derived 2D crater boundaries was subsequently developed and applied. 
Using this tool, the best possible 3D profiles for various scale ranges were then extracted and cross verified using the 2D crater data. 
The algorithms which are demonstrated in this paper will provide a very powerful tool for future planetary studies. 
1. INTRODUCTION 
After more than four decades of research and manual efforts, 
only a few tens of thousands of the millions of craters on Mars 
have been catalogued, mainly those with diameters > 5 km. 
Automated techniques for crater detection and cataloguing are 
therefore necessary to take advantage of the vast quantities of 
remotely sensed data now available, especially now that 3D 
information is routinely available from the ESA Mars Express 
HRSC instrument (Albertz et al., Scholten et al., 2005). 
For the results of automated crater detection systems to be 
useful, high levels of accuracy must be achieved. This raises 
two immediate problems. Firstly, the accuracy of existing 
automated approaches is generally unsatisfactory and needs to 
be improved. Secondly, the lack of ground truth data makes it 
difficult to perform a meaningful and reasonably objective 
assessment. A further challenge is the development of a fully 
automated crater detection system capable of achieving 
accuracies > 95%. 
A number of databases of craters on Mars are readily available 
through the United States Geological Survey (USGS) website 
for the Planetary Interactive GIS Web Analyzable Database 
(PIGWAD) such as the Barlow catalogue containing 42,283 
craters of diameters > 5km. This was produced in the late 
1980s from Viking 1:2,000,000 scale imagery with ~231m/pixel 
resolution and is currently undergoing revisions to include Mars 
Global Surveyor and Mars Odyssey data (Barlow et al., 2003). 
Also available through PIGWAD are the Roddy catalogue of 
4,300 craters > 10km diameter, with morphological details; the 
Kuzmin catalogue and the Costard catalogue of 2,600 craters 
with fluidized ejecta blankets. All have crater diameters >5km. 
The most important application field for reliable crater 
cataloguing and counting (using the Size Frequency 
Distributions) is the surface dating of planets. For example, 
surfaces on Mars have been subjected to a greater degree of 
modification than the Moon due to processes unique to Mars, 
such as water erosion. In addition to volcanic lava flows, there 
are areas of mobile sand dunes that obliterate smaller craters 
faster than larger ones, there is evidence of large-scale fluvial 
flows and ice movement with clear examples of surfaces that 
have been buried and/or exhumed. Whilst increasing the 
complexity, these dynamic processes can provide valuable 
information when analysed together with crater size-frequency 
distributions. Hartmann (1966) described how the age of a 
surface derived from crater counts is dependent both on the rate 
of crater formation and on the length of time that craters of that 
size last before evidence is lost due to erosion etc., Hartmann 
coined the term “crater retention age” as the average time 
interval that craters of a given size are retained on a surface. 
There is general agreement that the Solar System experienced 
very high impact rates during a period from around 4.6 to 3.8 
billion years ago, known as the “heavy bombardment”, which 
then rapidly lessened and became relatively stable. This 
relatively stable cratering rate is the basis for surface dating 
techniques which use the distribution of sizes and counts of 
craters. 
However, there are two problems to employ existing catalogues 
or in future more detailed catalogues for Martian chronology 
and geomorphological research. First of all, these catalogues 
focus on the larger craters, generally with diameters > 5km, 
whilst the target diameter for the application of geological 
research is mainly from around a few hundred metres up to a 
few km, essentially covering the size frequency distribution 
below the Barlow catalogue. 
The other problem is the deficiency of 3D information on 
craters. For research on erosional process over impact craters, 
the DTMs of corresponding crater is absolutely essential. The 
other need for 3D information is to enable the extraction of the 
depth-diameter ratio or the bottom shapes of crater. Secondary 
craters affect the dating of surfaces using crater counts since a 
single large primary impact can cause a great number of 
secondaries which will artificially increase the count and thus 
increase the age estimate. 
* Corresponding author