Ramon Gonzalez, Paramsothy Jayakumar, Karl Iagnemma
Journal of Terramechanics, Volume 68, December 2016, Pages 23-35, ISSN 0022-4898, http://dx.doi.org/10.1016/j.jterra.2016.09.002.
http://www.sciencedirect.com/science/article/pii/S0022489816300465
Abstract:
This paper presents an efficient method for increasing the accuracy of one key step regarding the process of determining a mobility map. That is, the interpolation of the original Digital Elevation Model (DEM) to a finer resolution before running multi-body-dynamics simulations. Specifically, this paper explores the use of fractal dimension and elevation range metrics for increasing the accuracy and reducing the computation time associated with the spatial interpolation ordinary kriging method. The first goal is to ensure the stationary variogram requirement. The second goal is to reduce kriging error or variance in the new predicted values. A novel segmentation-based approach has been proposed to divide the regions of interest into segments where stationarity is ensured. Empirical investigation based on real DEMs indicates the generality of the segmentation approach when natural and man-made terrains are considered. The proposed method leads to a more efficient computation burden and to more accurate results than the traditional approach.
Keywords: Mobility prediction; NATO Reference Mobility Model (NRMM); Variogram; K-means; Man-made environment