Masonry heritage buildings are typically vulnerable under seismic forces, due to the original design based on gravity load and, at the same time, the long-term decay effects. Nowadays, the computation of the vulnerability index is crucial for seismic evaluation and conservation strategies. An accurate geometric survey is a key-step for the learning process, especially in geometry-governed structural elements (e.g. in vaults, members with variable cross-section, etc.). In the case of huge building, scaffolding is commonly assessed in order to touch the structure for measuring or sensors installing (e.g. laser scanner). Minor damages are accepted when scaffoldings are fixed to the structure even in case of cultural Heritage. In the present study an innovative drone-based technique for geometric survey is proposed, it is aimed to be fully non-destructive. A drone is equipped with a high-resolution camera (3D rotating) and GPS-system for the spatial positioning. By reaching a sufficient number of photos, the whole surface of the building can be covered. The database is then processed in order to compute a points cloud. Consequently, the points are linked and the wireframe model is reached. Finally, surfaces are generated in between the edges and the meshed model is made. The procedure is intended to be reproduced over the time for monitoring potential crack or local instability. In order to demonstrate the validity of the proposal, a pilot building was selected. It is a masonry church with a large dome on the top (about 15 m diameter) and the variable thickness of the walls (about 50%). The indoor and outdoor photos were acquired by drone. Finally, the solid model was met by means of a specific designed Grasshopper-based code. In addition, the solid model was imported in a finite elements habitat, ready for the structural analysis.