|
Jose Saraiva, Instituto Superior Tecnico (Portugal)
Lourenco Bandeira, Instituto Superior Tecnico (Portugal)
Pedro Pina, Instituto Superior Tecnico (Portugal)
Catarina Barreira, Instituto Superior Tecnico (Portugal)
|
|
With the current effort to understand the evolution of Mars, there has been a marked increase in the number of images of its surface acquired by the automatic probes that orbit the planet. Furthermore, these images have a much better spatial resolution than those available before the turn of the century. Thus, the amount of information about the geological processes that took place on the planet and influenced the present morphology of its surface has grown in an almost overwhelming manner, which is extremely demanding on those that interpret the images. The attention they give to particular types of features can easily lead to others being discarded. However, certain features such as impact craters can be detected and (to some extent) characterized by automated procedures, thus freeing human operators to pay attention to other characteristics of a planetary surface.
In recent years, a number of procedures for the automated detection of impact craters have been proposed, but none seem to have met a welcome reception from those more deeply involved in using those features for different goals (mainly chronological studies). The reason for this, apparently, stems from a deep mistrust of the ability of mathematical procedures to clearly distinguish between craters and other quasi-circular shapes on a planetary surface. In fact, this is a problem that can easily affect analysis by human operators, and that only experience can resolve, though not eliminate completely. Thus, it is perfectly acceptable to have a degree of error (and thus, some uncertainty) in manual counts of craters.
We developed an approach to the problem of automated detection of impact craters that has yielded results that can be compared with traditional (manual) methods, not only in the actual detection and location of impact craters on the surface of Mars, but also in the conclusions that can be derived from this type of data concerning crater densities and age of surfaces.
This methodology has been tested in images of several areas of the surface of Mars, and also in images acquired by different cameras, with diverse spatial resolutions. There is plenty of room for improvement, and further developments of the methodology can and will be performed, but we present some interesting results that point to the usefulness of the application of such an automated procedure. When faced with large numbers of high spatial resolution, human operators can be greatly helped by a first definition of the impact craters present that does not show obvious errors. The job of confirming and detailing small scale craters would take a much smaller toll on the time and need for intense concentration of the researchers.
|
