Interpolation-Based Modeling for Automatic Digital Elevation Model Extraction from LiDAR Data.

Existing algorithms for terrain extraction from LiDAR require data to be structured in either a regular grid or triangulated irregular network (TIN) format, and they involve two-step processes; namely filtering and interpolation. This paper introduces a new terrain model based on reversing common interpolation techniques. A new method is presented to extract terrain directly through the inverse interpolation model in a single step. In addition, no raster or TIN format is required to represent the LiDAR data. Instead, points can be processed sequentially as obtained by the laser scanner. Furthermore, additional constraints for handling missing data, which occurs within waterbodies or occluded areas, are introduced in this paper. Several examples of extracted terrain in a variety of scenarios are presented and compared to results from existing algorithms both numerically and visually by inspecting vertical profiles across the data. The new terrain model was found to be in very close proximity to that obtained using existing methods but the former provided a smoother version of the terrain. [ABSTRACT FROM AUTHOR]