Classification of Martian Terrain Using Automated Discovery of Structure Algorithm Applied to Digital Topography

Details Classification of Martian Terrain Using Automated Discovery of Structure Algorithm Applied to Digital Topography Classification of Martian Terrain Using Automated Discovery of Structure Algorithm Applied to Digital Topography

Dec 2003

adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2003agufm.h11f0906v&link_type=abstract

American Geophysical Union, Fall Meeting 2003, abstract #H11F-0906

Mathematics

Probability

1824 Geomorphology (1625), 3210 Modeling, 6225 Mars

Scientific paper

The morphology of Martian landscape is of great interest because it helps to identify physical processes responsible for the observable topography. Traditionally, the descriptive method, applied to imagery data, has been used to study and categorize different types of Martian landscapes. We are developing a complementary approach, wherein a landscape is classified by a computer algorithm on the basis of digital topography provided by the Mars Orbiter Laser Altimeter data. We have adopted the automatic discovery of structure (ADOS) methodology, an unsupervised learning technique that classifies the data by grouping together similar cases. We use probabilistic algorithm that groups cases into classes by modeling each class through probability density function. Each case has a probability of class membership and is assigned to the class with highest posterior probability. The optimal number of classes is determined by cross-validation. The ADOS algorithm is applied to group pixels in a digital elevation model (DEM) of Tisia Valles, a typical Noachian Martian surface located at 46.13E, 11.83S. This terrain is heavily cratered, and shows presence of channels. An auxiliary DEM of the same size is calculated to contain an elevation field modified to make the landscape drainable. The DEM has 163240 pixels, each pixel carries its local topographical information encapsulated in a list of six quantities (h, δ h, s1, s2, a1, a2) which we call a topography descriptor. The components are: elevation, elevation difference between drainable and original DEMs, slopes in original and drainable DEMs, and contributing areas in original and drainable DEMs, respectively. Euclidean metric in space of topography descriptors is used to measure the ``closeness'' between pixels. The algorithm partitioned the pixels into 12 well-separated classes. Comparison of spatial distribution of these classes with visual rendering of digital topography reveals a geomorphic significance of obtained classification. Interiors of craters, ridges, inter-crater planes and channels are separated into different classes. Some subtle differences between otherwise similar terrain are picked up by our classification. Four classes represent crater interiors; they discriminate between different crater depths. Four classes represent inter-crater plains, they differ by actual elevation. Three classes represent ridges, they discriminate between different slopes. Finally, a single class represents channels. Using this classification we have constructed a thematic map of the Tisia Valles region that portrays spatial relations between various geomorphic features.

Affiliated with

Also associated with

No associations

Rating

Classification of Martian Terrain Using Automated Discovery of Structure Algorithm Applied to Digital Topography does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.

If you have personal experience with Classification of Martian Terrain Using Automated Discovery of Structure Algorithm Applied to Digital Topography, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Classification of Martian Terrain Using Automated Discovery of Structure Algorithm Applied to Digital Topography will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFWR-SCP-O-1424691