Rilievo e modellazione 3d di un sito archeologico tramite strumentazione UAV: confronto tra approccio closed- ed approccio open-source

The aim of the project was to realize a photogrammetric model of the excavation of the Poggio alla Regina Castle, useful for creating a detailed mapping of the georeferenced entities present on the ground. In addition, the generation of a stereoscopic model of the archaeological excavation allowed to generate a detailed point cloud from which it was possible to derive an accurate 3D model, useful to make measurements with an high level of accuracy. The drone survey was planned in the laboratory; it consisted of two different flight plans at two different heights, 35 and 50 meters above the ground surface. Great attention was kept on the percentage of overlapping region between single frames and between adjacent strip. In order to obtain the best results it was decided to have a 70% of longitudinal overlap and 50% of transverse overlap (sidelap). The first stage in the field was the positioning of a sufficient number of Ground Control Points (GCP) on the ground. The 3D coordinates of every GCP were then acquired via Differential Global Positioning System (DGPS) techniques, using the stop-and-go method with an observation time of 2 minutes at each point. After the topographic survey, an AscTec Falcon 8 vehicle, owned by the Geographike Ltd. Company and equipped with a Sony NEX-5N camera (focal length fixed to 24 mm), was used to perform 4 different flights. The first flight was carried out as previously specified, taking images of the area at two different heights, 35 and 50 m. Moreover, a second flight was performed at a lower level, 20 m. The third flight was carried out in order to take a complete “panorama” of the area, while the fourth flight was dedicated to a video record. The entire survey was completed in one day, highlighting the short operational time needed. The data were then post-processed following two different approaches. It was initially used a closed-source software (Agisoft PhotoScan), which led to the creation of the georeferenced point cloud, DEM, orthophoto, and 3D model of the area of interest. The software uses a powerful algorithm that allows to orient the acquired frames. The advantage of this approach is that the processing operations are fully automatic and can be entirely completed using a single software. In the second approach the same outputs were realized using open-source software, such as Pyton Photogrammetry Toolbox for the orientation of the frames and the realization of the point cloud, Sfm_Georef for the georeferencing, and MeshLab for the creation of the 3D model. The final outputs obtained from the two different approaches were then compared in terms of quality and accuracy. This was done in a GIS environment (ESRI ArcGIS), where some points corresponding to the GCP placed on the ground were used to assess the models accuracy. In conclusion the advantages and disadvantages of each approach were discussed, taking into account costs, time required for the elaborations, necessity of powerful computers and level of quality of the obtained products.