TERPOLI (2007-2010)
Automatic recognition and characterization of polygonal terrains on Mars
Activities
The main tasks of this project were the following:
The main tasks of this project were the following:
- Creation of image bank - A large data base of images of the surface of Mars in which polygonal terrains can be seen was created: from about 20,000 images inspected, more than 1500 contained polygonal networks. These Martian images were captured by MOC, CTX and HiRISE cameras.
- Visual evaluation of selected areas - Two main results were obtained from the execution of this task: (i) Definition and application of criteria for the definition of classes of polygonal patterned terrains, according to relevant factors such as dimension and location: images with a spatial resolution better than 6 metres/pixel located above 50º in latitude (N and S) weer selected. The second is a set of results concerning a number of areas selected because of the presence of polygonal patterned terrains, and which will illustrate and represent the different types previously defined. These results, which included the manual delineation of the polygons, were taken as ground-truth for the validation of the automated procedure for identification and characterization of this type of terrains, which was the main objective of this project.
- Automated identification of polygonal patterns - A working version of an automated methodology for the identification of polygons in images of the surface of Mars in which they are present was developed. This methodology, mainly based on mathematical morphology, produced precise and accurate results, namely by achieving high recognition rates (the correct identification of pixels as being part of the dividing lines between adjacent polygons was normally above 90%) and very low rates of false positives (always below 10%).
- Geometrical and topological characterization of polygonal patterns - This was achieved through the collection of a number of characteristics of the previously identified polygons, related to geometric and topologic features, such as: (i) Geometric: area, perimeter, mean axis, orientation - Topological: number of polygons with “i” sides, adjacencies between “i” sided and “k” sided polygons (“i” different of “k”).- Geometric-topological: distribution of surface areas of polygons versus topology (number of sides and adjacencies). Some classic laws (Lewis, Desch and Aboav-Weaire) were experimentally verified for a large and diversified type of Martian networks. A first attempt ot classify the Martian metwirks into types using the geoemtric and topologic features was performed, some clusters were obained. Some fast algorithms to deal with the topologic features were also developed.