Construct massive terrain models with SCALGO Model

A Delaunay triangulation of the (2D projection of the) input points set is used to construct the TIN because it attempts to avoid skinny triangles by maximizing the minimum angle of all the triangles. The triangulation is constructed on all the input points to avoid artifact such as those that can e.g. be introduced when using tiling. If the input point set is given in the LAS file format, where each point is assigned a class, then the module supports constructing the TIN on specific classes.

After constructing a terrain model one often is interested in adding or removing certain polygonal features (such as buildings), or modifying features that are already in the model (such as to make sure that water bodies are completely flat). The below figure illustrates a terrain model built with the TIN construction module where building have been added and bridges removed (to perform hydrological analysis, e.g. with the SCALGO Hydrology package) using the SCALGO Utility Burn module.

Example of how an extra point in the point set above can introduce three large triangles. These triangles can be ignored in the interpolation since they have large side lengths.

As illustrated in the example above to the right, no interpolation is performed in (the non-shaded) cells where the center is not contained in a triangle, that is, elevation values of raster cells outside the convex hull of the input points are undefined (given a nodata value). As illustrated in the figure on the right, it is sometimes desirable to ignore large triangles in the interpolation step. For instance, large lakes in the terrain may result in large triangles in the interior of the model, and inlets and fjords may result in large triangles along the boundary of the model. Interpolating over these triangles will often not yield appropriate elevation values. Optionally, it is therefore possible to ignore all triangles where the length of any of the three sides exceeds a given threshold.

Example of a raster where a cell is colored red if the distance to the nearest input point is longer than 1 meter is overlaid with an orthophoto. It is clearly seen that the input lacks points on buildings and vegetation.

As illustrated on the figure on the right, the large valued cells in the point coverage raster correspond to regions that are sparsely covered with input points and thus are the regions where an interpolated terrain model is less reliable.