Automatic Recognition of Impact Craters on the Martian Surface from DEM and Images - Planetary Exploration and Science: Recent Results and Advances - page 105

of the highest value of H(C) within the group is considered the best one for the
positive identification of the craters, while the remaining ones are discarded without
becoming false negatives.

6.2.2 Recognition from DEM

All the image-based crater-detection approaches discussed above involve compli-
cated, multistep algorithms to combat inherent limitations of imagery data. On the
other hand, there are many factors, such as illumination effects, that increase the
possibility of the biased estimate of detection results. Furthermore, it has some
fundamental limits because of the 2D imagery data. Imagery data is well suited for
human visual interpretation but ill-suited for automated processing because images
are skewed representations of the landscapes they portray. However, we can easily
obtain DEM data which are more direct and well-suited descriptors of planetary
surface than images. Interestingly, no complete DEM-based crater-detection algo-
rithms have been developed in the past. The less interest in DEM-based algorithm
probably results from the scarcity and limited resolution of planetary topography
data.
Currently, the available Martian DEMs are constructed from the Mars Orbiter
Laser Altimeter Mission Experiment Gridded Data Record (MEGDR) with a
resolution of 1/128 ı or 500 m at the equator. It has limited the size of crater that
can be detected to 3-5 km at most in diameter because of the coarse resolution of
the DEM. This is on par with the lowest size of craters in the Barlow catalogue and
small enough to extract a catalogue of scientific significance, and it has not met the
increasing requirement in navigation or research of Martian surface history. In the
near future, significantly higher-resolution DEMs of selected portions of Martian
surface will be compiled from the Mars Express high-resolution stereo camera
data which will greatly improve the ability of detecting the small size craters and
contribute to our further research.
Detecting craters from DEM is, in principle, more convenient than from visual
images because craters are landforms which can be calculated as terrain attributes
to be distinguished from other morphological features. As the topographic data can
be more easily available, more scientific teams pay attention in this field and have
some basic tests to find an effective algorithm. In this section, we will introduce two
new approaches in details.

6.2.2.1 Detecting Craters by Hough Transform

Preprocessing the DEM

A DEM is a raster dataset, where every pixel is assigned an elevation value and
labeled by a set of coordinates with x and y . Profile curvature, which is the curvature