We propose a novel method for automated classification of the Martian landscape into constituent landforms.
We construct a tri-feature digital grid that stores elevation, slope and curvature information for every pixel in a site. 
This is similar to a multi-spectral image. The grid is then segmented into fragments which are homogeneous with respect to these 
features.Also a shape parameter is calculated for each fragment. Each fragment is then represented by a feature vector containing 
elevation, slope and curvature (averaged over the constituent fragment pixels) and the shape parameter. These feature vectors, 
thus generated are then used for landscape classification.