This project is about DTM generation based on a photogrammetrically produced point cloud. The project originates from a desire to improve the DTM algorithm proposed in the previous project “Terrain Modelling – DTM Generation using UAVs” by Matthesen and Schmidt (2014).

In the previous report, a number of problems were experienced. Based on investigations of these problems, a new and optimized algorithm has been proposed. In the proposed algorithm, two steps are used to remove the non-terrain points.

The first step uses a surface-based filter to remove the big clusters of non-terrain points like e.g. houses and trees. The used order of the fitted surface polynomial is automatically adjusted according to the terrain.

The second step of the algorithm removes the remaining non-terrain points by using a slope-based filter which automatically adapts itself according to the slope.

The optimized algorithm is implemented in a GUI program written in Python and Cython. Using this program, the algorithm was applied to different test-areas, and the quality of the algorithm could be assessed.