4.1.2. San Cono Bridge

San Cono bridge spans the Bianco river located in the municipality of Buccino, in southern Italy (Figure 3a,b). As reported by the inscription on the bridge, the construction of San Cono bridge can be dated to the Augustan age (Figure 3c).

**Figure 3.** San Cono bridge: location (**<sup>a</sup>**,**b**) and panoramic images of masonry bridge (**<sup>c</sup>**,**d**).

Originally, the bridge had a pronounced donkey-back profile with two shoulders and a steep slope at the ends and a pylon with a triangular rostrum [20]. Now, the current shape of the bridge is incorporated into a new bridge, which in 1872 levelled the road and widened the site (taking it from 3.20 m to 6.45 m), covering it, so as to leave only the original arches visible, below the new ones. In this way, the intervention represented an exceptional example of respect for the ancient monument. As for the bridge architecture, it has two spans of unequal light, for a total length of 40 m. Part of the ancient arches can still be seen below the nineteenth-century one, which changes its profile. The central round arch has a light of 17.3 m and at the base there are five projecting brackets with three others at a higher altitude to complete the support of the rib; the minor arc has a light of 5.9 m with three shelves.

The original vestments of the tympani were in square work; today they are inserted in the new 19th-century vestments, with an upper parapet that modifies the original donkey back profile [20] (Figure 3d).

#### *4.2. Three-Dimensional Point Cloud of San Nicola in Montedoro Church*

#### 4.2.1. Three-Dimensional Survey of the Church

The survey of the church was carried out through the use and integration of active and passive sensors, terrestrial and aerial. In particular, the external façade was surveyed using a TLS, the inner part using a digital single-lens reflex (DLSR) camera with fish-eye lens and the upper part of the building (i.e., the roof and other architectural elements not visible through a terrestrial survey) through the use of a camera mounted on a UAV platform.

Before performing the surveys with photogrammetric techniques and laser scanners, a survey with a total station was performed. The survey was carried out by TS30 Leica Geosystems. This total station allows discrete points to be acquired with an angular precision of 0.5" (0.15 mgon) and to acquire distance with prism (precision of 0.6 mm + 1 ppm) and without prism (2 mm + 2 ppm).

In this case study, the survey was carried out by two base stations. In this way, it was possible to obtain horizontal and vertical angular observations of the ground control points (GCPs). The GCPs, inside and outside the building, were chosen so as to be easily recognizable even on the image (Figure 4). The post processing of the data was carried out in LGO (Leica Geo O ffice) developed by the Leica Geosystem company.

**Figure 4.** Images of some ground control points used for the georeferencing of the point cloud.

4.2.2. Survey of the Terrestrial Laser Scanner of the External Part of the Structure

Regarding the generation of the model for the external part of the church, the survey was carried out by a terrestrial laser scanning survey. In this case study, FARO FocusS 350 instruments were used because specially designed for outdoor applications. HDR imaging and HD photo resolution (overlay up to 165-megapixel colour) ensure true-to-detail scan results with high data quality (distance accuracy up to ± 1 mm). The main features of this scanner are summarized in the following Table 1:

**Table 1.** Main Features of the FARO FocusS 350 terrestrial laser scanner (TLS).


In order to cover the entire external surface of the church, three acquisition stations were built.

The post-processing of the TLS scans was performed in Autodesk Recap software. This software, where the word "Recap" stands for Reality Capture, allows a fully automatic recording of the scans. In the case the procedure is partially successful, the software allows manual identification of targets and natural homologous points, to reduce distance among contiguous scans, improving their alignment using the iterative closest point (ICP) algorithm [21].

4.2.3. Unmanned Aerial Vehicle (UAV) Photogrammetry to Obtain the 3D Point Cloud of the Upper Part of the Church

The aerial survey was carried out using a Parrot Anafi, a UAS (unmanned aerial system) quadcopter equipped with a Sony Sensor® 1/2.4" 21MP (5344 × 4016) CMOS (complementary metal-oxide semiconductor), which allows obtaining, thanks also to a 3-axis stabilizer, clear and detailed images (Figure 5a). The distance between the UAV and the building was really close due to the presence of many obstacles in the old town where the church is located. Consequently, the images were acquired with high geometric resolution (Figure 5b). In any case, the photogrammetric survey was carried out with a high degree of overlap between the images. In addition, by varying the tilt angle of the camera, it was possible to acquire images of every part of the building. In this way, it was possible to build a network of the 97 images with a high degree of overlap and convergen<sup>t</sup> image configuration (Figure 5c).

**Figure 5.** Image acquisition by unmanned aerial vehicle (UAV): (**a**) Parrot Anafi and Parrot Sky-controller; (**b**) acquisition step; (**c**) an aerial image acquired by UAV platform.

Taking into account 5 GCPs, the root mean square error (RMSE) for spatial coordinates, evaluated on the cameras used in this dataset, was of 0.009 m; in particular, this RSME refers to the georeferencing process of the images and not to the resolution of the model. In addition, the modelling of the roof was obtained through the insertion of its own and accidental load.

4.2.4. Photogrammetry of the Internal Part of the Structure Using a Fisheye Lens

For the interior of the church, since there are also frescoes of grea<sup>t</sup> historical and cultural value and considering the rather restricted environment, a photogrammetric survey was carried out using a Nikon D5000 DSLR camera with a calibrated fisheye lens (focal length 10 mm). The fisheye is a wide-angle photographic lens that allows a wide scene to be observed. This type of lens has been used successfully in the photogrammetry field, as shown in Kannala and Brandt, 2006 [22], especially in narrow spaces. In self-calibration mode, the dataset of the 22 images was processed in Agisoft Metashape software. The total error, i.e., standard deviation evaluated on 6 GCPs, was 0.003 m.

Considering the high value of the frescoes and the architecture of the small altars inside the structure, orthophotos of each single façade and floor were taken. In order to carry out this task, it was necessary to build a mesh of the interior of the structure. Subsequently, identifying the planes of the single façade, the orthophotos with a geometric resolution of 0.1 mm of the interior of the church were built (Figure 6).

**Figure 6.** Orthogonal projection of the inside of the structure: orthophotos (in very high resolution) of the single façade and floor of the church.

4.2.5. Merging of the Datasets (Point Clouds)

Through the survey activity and post processing of the data obtained either with IBM or RBM methods, it was possible to obtain three datasets, as shown in Table 2.

**Table 2.** Point Clouds Obtained in the Several Datasets.


The several point clouds were merged in a single point cloud on the base of common point. This task was carried out in 3DF Zephyr environment, which is a commercial photogrammetry software, developed and marketed by the Italian software house 3DFLOW. A representation of the whole structure according to point cloud is shown in Figure 7.

**Figure 7.** Three-dimensional (3D) point cloud of San Nicola in Montedoro.

#### *4.3. Three-Dimensional Point Cloud of San Cono Bridge*

In order to build the 3D model of the bridge, the photogrammetric survey was divided into an aerial and a terrestrial one. Taking into account the scale of representation (SR) and the aim of the project, a Ground Sample Distance (GSD) equal to 1 cm was chosen as reference for the survey. The terrestrial survey was carried out in order to survey the lower part of the bridge using a Canon EOS 100D DSLR camera (Charged Coupled Device -CCD size = 4.29 μm) with a focal length of 18 mm. A total amount of 400 terrestrial images was acquired. As regards the aerial survey, this was carried out using a UAS Xiaomi Mi 4K, a multi-copter rotary wing weighing less than 1.5 kg and whose declared maximum speed is 18 m per second (about 65 km/h). This UAV was developed and produced by Flymi, a company of Mi Ecosystem. The photogrammetric features of the camera mounted on UAV platform were: CCD size = 4.29 μm and focal length of 3.5 mm. The aerial survey was designed using a software called Mission Planner, which is developed by Oborne for the open-source APM autopilot project. The flight plan was designed with the following characteristics [23]: 80% longitudinal (end-lap) and 60% transversal overlap (sidelap). In addition, flight lines (FLs) inclined at 30◦ and 45◦ were designed in a direction longitudinal to the bridge in order to increase the rigidity of the aerial photogrammetric block and, at the same time, to increase the redundancy of information with the data obtained from the terrestrial survey. In total, 285 images were taken during the aerial survey.

The post-processing of terrestrial and aerial images was carried out using Agisoft Metashape software. In this case study, two separate chunks were built: one involving aerial (UAV) surveying and another involving terrestrial surveying. To evaluate the quality of image matching (alignment step), the number of the projections and the error achieved on the single chunk were taken into account. Table 3 shows the high quality of the image matching and, consequently, the correctness in the phase of acquisition, for both the aerial and the terrestrial surveys.

**Table 3.** Report on image matching for the two datasets.


According to the photogrammetric pipeline, a dense point cloud was built for both datasets. Consequently, in order to obtain the model of the bridge under investigation, it was necessary to integrate the two datasets on the basis of common points. In total, the final 3D point cloud consisted of approximately 8 million points (Figure 8). Subsequently, the model was scaled using 12 Ground Control Points obtained through a traditional topographic survey.

**Figure 8.** Three-dimensional point cloud of San Cono bridge (visualization in Agisoft Metashape software).

#### *4.4. Three-Dimensional Reconstruction of the Models*

The point cloud obtained from the geomatics surveys must be classified in objects which the structure under examination consists of. The processes necessary to perform this task must take into consideration several parameters, such as noise, occlusions, the association between faces of neighbouring objects, etc. We carried out this task in Rhinoceros software because it has more tools and plug-ins for 3D modelling. The key point of this software application is the possibility of generating a profile of the structure and, especially, to build a surface that can be adapted to the point cloud obtained in geomatics surveys. Once the point cloud was imported into Rhinoceros, it was possible to reanalyse it using the Arena plug-in. In this way, the density of the points of the PC was decreased and, consequently, it was possible to assess if there were any holes in the 3D model. Within the Rhinoceros software, the tools available to users are quality, point size and the visual analysis tools (render, ratio, opacity). This allowed for editing the point cloud of the structure. Subsequently, the point cloud was dissected into several planes in space. This operation allowed sections in strategic points of the structure, such as the arches of the bridge (Figure 9), to be performed.

**Figure 9.** Scheme of the position of the sections.

The plug-in allowed saving the sections in a specific layer. As a result, the sections were displayed as "construction plans" (see Figure 10).

Sections that are transverse and longitudinal to the structure were used to create NURBS. Using the EvoluteTools PRO plug-in, it was possible to generate NURBS surfaces (Figure 11a). This plug-in allowed us to shape NURBS surfaces on objects of the structure, exploiting both the sections and the

point cloud through an appropriate algorithm developed within this plug-in. For example, the bridge pillar was modelled using an adaptive NURBS (Figure 11b).

**Figure 10.** Construction planes within Rhinoceros software.

(**b**)

 **Figure 11.** Adaptation of the non-uniform rational B-spline (NURBS) to the existing surface of the masonry bridge: detail of the arches (**a**) and of a pylon (**b**).

Of course, the time of the clustering task was related to the complexity of the structure. In this way, it was possible to create surfaces that represent the elements of the structure (vault, stack, retaining walls and superstructure of the bridge), as shown in Figure 12a. Using the same procedure just described for the masonry bridge, it was possible to build a 3D model of the San Nicola in Montedoro church too (Figure 12b).

**Figure 12.** Three-dimensional model created in Rhinoceros environment: exploded structural elements of the bridge (**a**); 3D model of the San Nicola in Montedoro church (**b**).

Lastly, thanks to the development of the Grasshopper plug-in, it was possible to model similar structural elements (or parts of them) in 3D. Thus, it was possible to parameterize both from the geometric point of view and from the point of view of the type of material. For example, Figure 13 shows the parameterization of the arch of the bridge using the tools developed in Grasshopper.

**Figure 13.** Three-dimensional model of the masonry bridge using Grasshopper software.

## *4.5. Building Information Modelling (BIM)*

Many commercial BIM software products are available on the market. One of the most efficient is Autodesk Revit. The original software was developed by Charles River Software, founded in 1997, renamed Revit Technology Corporation in 2000, and acquired by Autodesk in 2002. Autodesk Revit allows users to design a building and structure and its components in 3D, annotate the model with 2D drafting elements and access building information from the building models database. Modelling in the BIM environment of the two case studies was carried out using Autodesk Revit software.

In both case studies, the resulting mesh surface obtained in Rhinoceros software in 3D ACIS Modeler (ACIS) format (\*.sat) was imported into the BIM Revit software. In this way, the surface created can be quickly opened by the BIM software and can be easily manipulated with rotations and translations. The high detail of the polysurface allowed the precise determination of the levels for the creation of BIM objects. Screenshots of the modelling and managemen<sup>t</sup> of the information in Revit software, both of the masonry bridge and of the church, are shown below (Figure 14 a,b).

(**b**) 

**Figure 14.** Visualization of the structures in Revit environment: San Nicola in Montedoro church (**a**) and San Cono masonry bridge (**b**).
