ANNOTATION
This Bachelor Thesis deals with determining which methods of adjusting Digital Terrain Models are optimal for the needs of aerial imaging orthogonalization. The aim of the thesis is a creation of filters and filter strategies in DTMaster software, which is included in the Trimble Inpho program pack. The key function of filters is to automatically go through point clouds created by the aerial imaging and detect from them all buildings, vegetation and gross errors. The goal is to mark all objects which are located above the ground while the non-classified points are taken to be the ground. This is achieved by forming a filter strategy. Such strategy is the result of combining individual filters. Points making out the ground are then used to create a Digital Terrain Model, hence the creation of the orthomosaic. The better the ground points classification, the better the quality of the resulting outputs.
Two case studies were chosen for the testing of filters. One provided images taken by an Unmanned System, whereas the other provided them by aerial imaging. Both case studies differ in Ground Sample Distance, diversity of vegetation and height of buildings. The accuracy of filters was tested by setting various parameter values. Although these values are already recommended, each case study is special and requires certain adjustments. Therefore, additional settings, which were suggested by available literature and own measurements, were also tested.
The accuracy of the automatic filtrations was then compared with reference data that were made by manual filtration. The comparison was done in two ways. The first one was statistical, and it was based on the calculation of referenced and evaluated distance of points within the point cloud. The second way was visual. It consisted of comparing accurate projections of objects (straight edges of buildings, line elements as pathways and vegetation) automatically filtered orthomosaics to manually filtered orthomosaic. The outcome of this thesis are the best rated filter strategies, together with recommendations regarding their further practical utilization.