Polygonal simplification and its use in DEM generalization for land surface segmentation

Abstract

We introduce a simplification method based on irregular triangular networks to create a generalized digital elevation model (DEM) for land surface segmentation. The quadric error metrics simplification (QEMS) algorithm was used to generalize the DEM as a method that follows the theory of optimal triangle for (land) surface representation. Its use for land surface modeling is novel, however, it is well known in computer graphics. Comparison of QEMS with the maximum z-tolerance method shows priority in terms of root mean square error. Numerical expression of the value concentration of the third-order variables (curvature changes) around zero (K0) was used as an indicator of elementary form representation in generalized DEMs. K0 was calculated for a set of DEMs at several generalization levels in two different territories, as well as for comparably generalized artificial surfaces. K0 values of real DEMs vary significantly with generalization levels and character of the landforms, and their maxima point to the generalization levels that best preserve the information about elementary forms of various hierarchical levels. Differences between K0 of real DEMs and artificial surfaces confirm the same.