ABSTRACT
This research proposes a methodology capable of automatically
and simultaneously detecting and delineating impact craters on the
surface of Mars. The proposal is based on mathematical
morphology fundamentals and digital image processing
techniques. The methodology is composed of four sequential
steps: (i) pre-processing and filtering to remove noise and enhance
feature edges, (ii) watershed segmentation and use of contour
dynamics strategies for crater detection and delineation, (iii) postprocessing
and detection refinement, and, finally, (iv)
performance evaluation. The dataset consisted of craters
represented by high spatial resolution images of the surface of
Mars, with 12.5 m/pixel resolution from the High Resolution
Stereo Camera (HRSC) and 1 m/pixel resolution from the High
Resolution Imaging Science Experiment (HiRISE). The results
point to overall true detection rates over 81% and false detection
rates between 10% and 13%. Therefore, we concluded that the
present approach contributes to solving a current problem in the
context of impact crater detection and delineation on the surface
of Mars.