Figure 1.
Test area: (a) The Rocca in the ‘Borgo Medievale’ and (b) The Valentino Castle.
Figure 1.
Test area: (a) The Rocca in the ‘Borgo Medievale’ and (b) The Valentino Castle.
Figure 2.
The main staircase in the courtyard of (
a) the Fenis Castle (AO) in 2004 and (
b) back to 1884 [
29]. (
c) The same scene in the courtyard of the Medieval Rocca in Turin.
Figure 2.
The main staircase in the courtyard of (
a) the Fenis Castle (AO) in 2004 and (
b) back to 1884 [
29]. (
c) The same scene in the courtyard of the Medieval Rocca in Turin.
Figure 3.
Terrestrial Laser Scanner (TLS) by FARO®: (a) The FARO® Focus 120 during the acquisition and (b) planimetric view of an outdoor coverage range of an X330 scanner.
Figure 3.
Terrestrial Laser Scanner (TLS) by FARO®: (a) The FARO® Focus 120 during the acquisition and (b) planimetric view of an outdoor coverage range of an X330 scanner.
Figure 4.
The ZEB Revo Real Time Mobile Mapping System (RT MMS) by Geo Simultaneous Localization and Mapping (GeoSLAM): (a) The data logger, the rotating head and the Go Pro mounted above and the IPad Pro employed for the real-time view during the acquisition phase; and; (b) Operators scanning the Great Salon with the instrument.
Figure 4.
The ZEB Revo Real Time Mobile Mapping System (RT MMS) by Geo Simultaneous Localization and Mapping (GeoSLAM): (a) The data logger, the rotating head and the Go Pro mounted above and the IPad Pro employed for the real-time view during the acquisition phase; and; (b) Operators scanning the Great Salon with the instrument.
Figure 5.
Different Unmanned Aerial Vehicle (UAV) platforms by DJI company and related sensors with captured images.
Figure 5.
Different Unmanned Aerial Vehicle (UAV) platforms by DJI company and related sensors with captured images.
Figure 6.
(a) The flights performed by Phantom 4 PRO (b) oriented UAV images block on the Rocca.
Figure 6.
(a) The flights performed by Phantom 4 PRO (b) oriented UAV images block on the Rocca.
Figure 7.
A view of the registered scans from (a) Rocca (n. 93) and (b) Castle (n. 82).
Figure 7.
A view of the registered scans from (a) Rocca (n. 93) and (b) Castle (n. 82).
Figure 8.
Example of SLAM-trajectory in the Rocca colourized by time: (a) Close loop with some roundtrip parts; (b) Roundtrip on the tower: going (blue) and return (red). (c) All of the reprocessed ZEB point clouds after the merge process, each colour corresponds to one scan, in blue the external one and (d) vertical section of the point cloud.
Figure 8.
Example of SLAM-trajectory in the Rocca colourized by time: (a) Close loop with some roundtrip parts; (b) Roundtrip on the tower: going (blue) and return (red). (c) All of the reprocessed ZEB point clouds after the merge process, each colour corresponds to one scan, in blue the external one and (d) vertical section of the point cloud.
Figure 9.
The ZEB survey in the Valentino Castle: (a) The plan view with the closed loop T2(τ), performed, before (blue) and after (range colours) the SLAM reprocessing by merge function correction; (b) the attributes related to ZEB-based point cloud processing, inside the GeoSLAM Hub platform. Along with an axonometric representation of the main floor in shaded view (I), there are normal mapping (II); SLAM quality condition during the scan (III); time-stamp information (IV). Each thematic result is associated with its own trajectory data as well.
Figure 9.
The ZEB survey in the Valentino Castle: (a) The plan view with the closed loop T2(τ), performed, before (blue) and after (range colours) the SLAM reprocessing by merge function correction; (b) the attributes related to ZEB-based point cloud processing, inside the GeoSLAM Hub platform. Along with an axonometric representation of the main floor in shaded view (I), there are normal mapping (II); SLAM quality condition during the scan (III); time-stamp information (IV). Each thematic result is associated with its own trajectory data as well.
Figure 10.
The desktop interface of the merge function in the GeoSLAM Hub software: in red, a single scan that is manually aligned, one at a time, to others, yellow. In the black box bottom right, the roto-translation matrix of each merged scan.
Figure 10.
The desktop interface of the merge function in the GeoSLAM Hub software: in red, a single scan that is manually aligned, one at a time, to others, yellow. In the black box bottom right, the roto-translation matrix of each merged scan.
Figure 11.
Statistical graphics of C2C comparison of ZEB and LiDAR point clouds on the Throne Chamber: mean and st. dev. distribution (a) before and (b) after noise filtering applied to ZEB data (C1-C2 cases).
Figure 11.
Statistical graphics of C2C comparison of ZEB and LiDAR point clouds on the Throne Chamber: mean and st. dev. distribution (a) before and (b) after noise filtering applied to ZEB data (C1-C2 cases).
Figure 12.
Extension of point clouds, UAV data coloured in blue, ZEB data in orange: (a) angular view of the external walls with the presence of vegetation and (b) vertical section of the Rocca.
Figure 12.
Extension of point clouds, UAV data coloured in blue, ZEB data in orange: (a) angular view of the external walls with the presence of vegetation and (b) vertical section of the Rocca.
Figure 13.
Complementary representation of C2C points distances between the two point clouds: (
a) ZEB distances on UAV data (min. 1.33 cm, Max. 9.27 cm) and (
b) vice-versa projection and values in
Table 6.
Figure 13.
Complementary representation of C2C points distances between the two point clouds: (
a) ZEB distances on UAV data (min. 1.33 cm, Max. 9.27 cm) and (
b) vice-versa projection and values in
Table 6.
Figure 14.
The benchmarking analysis on multisensors’ data of Fleur-de-lis Chamber in the Valentino Castle. (a) SfM points cloud; (b) LiDAR points cloud; and (c) ZEB points cloud.
Figure 14.
The benchmarking analysis on multisensors’ data of Fleur-de-lis Chamber in the Valentino Castle. (a) SfM points cloud; (b) LiDAR points cloud; and (c) ZEB points cloud.
Figure 15.
(a) The Roses Chamber vault and the (b) C2C distances analysis between LiDAR and ZEB.
Figure 15.
(a) The Roses Chamber vault and the (b) C2C distances analysis between LiDAR and ZEB.
Figure 16.
The two density analysis maps on the Roses Chamber in (a) LiDAR and (b) ZEB models.
Figure 16.
The two density analysis maps on the Roses Chamber in (a) LiDAR and (b) ZEB models.
Figure 17.
Morphological anomalies detected by the ZEB sensor: (a) the central part of the Fleur-de-lis Chamber vault in displacement map of T2–T1 and (b) in the Great Salon vault, with the comparison of isolines from the LiDAR DSM (white) and ZEB (black).
Figure 17.
Morphological anomalies detected by the ZEB sensor: (a) the central part of the Fleur-de-lis Chamber vault in displacement map of T2–T1 and (b) in the Great Salon vault, with the comparison of isolines from the LiDAR DSM (white) and ZEB (black).
Figure 18.
The n°4 points selected in (a) ZEB and in (b) TLS point cloud for coordinates’ extraction and finally the point was detected and placed on the (c) Digital Single Lens Reflex (DSLR) photo, for the images block-orientation.
Figure 18.
The n°4 points selected in (a) ZEB and in (b) TLS point cloud for coordinates’ extraction and finally the point was detected and placed on the (c) Digital Single Lens Reflex (DSLR) photo, for the images block-orientation.
Figure 19.
(a) The oriented images block in the Fleur-de-lis Chamber; (b) The Photoscan Pro Graphic User Interface (GUI) allows for importing external integrated point cloud for the fusion-based mesh triangulation.
Figure 19.
(a) The oriented images block in the Fleur-de-lis Chamber; (b) The Photoscan Pro Graphic User Interface (GUI) allows for importing external integrated point cloud for the fusion-based mesh triangulation.
Figure 20.
The ZEB-based surface reconstruction: (a) mesh processed with normal data computed with 15 mm radius and (b) computed with 5 mm; and, (c) with an HQ texture map applied on the (b) surface.
Figure 20.
The ZEB-based surface reconstruction: (a) mesh processed with normal data computed with 15 mm radius and (b) computed with 5 mm; and, (c) with an HQ texture map applied on the (b) surface.
Figure 21.
Graphic workflow of the hybridization of 3D models, deriving from integrated methods and fusion-based approaches. In grey, the acquisition methods, in green the achieved products with the possible connections. In the dashed box the fused and integrated results.
Figure 21.
Graphic workflow of the hybridization of 3D models, deriving from integrated methods and fusion-based approaches. In grey, the acquisition methods, in green the achieved products with the possible connections. In the dashed box the fused and integrated results.
Figure 22.
(a) A screenshot of the LiDAR points model managed in the Recap environment; (b) A screenshot from the video navigation achieved from LiDAR points cloud.
Figure 22.
(a) A screenshot of the LiDAR points model managed in the Recap environment; (b) A screenshot from the video navigation achieved from LiDAR points cloud.
Figure 23.
(a) The SLAM for full six-dimensional virtual reality/augmented reality (6D VR/AR) application running on an iPad; (b) Interactive model of the hall showing typical points enabling to reach additional information.
Figure 23.
(a) The SLAM for full six-dimensional virtual reality/augmented reality (6D VR/AR) application running on an iPad; (b) Interactive model of the hall showing typical points enabling to reach additional information.
Figure 24.
Relations between many deterministic factors turning around the Geomatics approach working in the cultural heritage domain.
Figure 24.
Relations between many deterministic factors turning around the Geomatics approach working in the cultural heritage domain.
Table 1.
Sensors employed in the acquisition phase and general overview of the amount of collected data for the entire project.
Table 1.
Sensors employed in the acquisition phase and general overview of the amount of collected data for the entire project.
Type of Survey | Systems | Sensors | Medieval Rocca and Hamlet | Valentino Castle Main Floor |
---|
Range-based | TLS | FARO® Focus3D X120 & X330 | 93 scans | 112 scans |
| MMS | GeoSLAM ZEB Revo RT+ZEBCam | 29 scans | 8 scans |
Image-based | UAV | DJI Phantom 4 Pro Obsidian DJI Mavic Pro Platinum DJI Spark | 1919 images | 264 images |
| DSLR | Sony ILCE 7RM2 Canon EOS 5DS R | 2455 images | 1942 images |
| Low-cost sensors | DJI Osmo+ GoPro Fusion 360 | 15 videos | 10 videos |
Topography | GNSS | Geomax Zenith 35 | 18 vertices 145 targets | 17 vertices 162 targets |
| TS | Geomax Manual TS Zoom30 Pro | | |
Table 2.
UAV acquisition specifications.
Table 2.
UAV acquisition specifications.
Specifications | Phantom 4 Pro Obsidian |
---|
Height flight | 50 m |
Area covered (Area of Interest AoI) | 0.106 km2 (4000 m2) |
Strips | 1 circular (oblique), 2 nadir |
N° of images | 137 |
Time | 45 min |
GSD (cm/px) | 2.34 cm |
N° of GCPs/CPs | 20/8 |
Pt density (AoI) | mean 1800 pt/m2 (17,500 pt/m2) |
Table 3.
Light Detection and Ranging (LiDAR) acquisition specifications.
Table 3.
Light Detection and Ranging (LiDAR) acquisition specifications.
| Environmental Features | Scans | Registration/Georeferencing |
---|
| n° Rooms | Surface | Volume | Time | n° | n°/Room | ICP | Targets-Based |
---|
(m2) | (m3) | (h) | Mean Error (mm) | n°CPs | St.dev (mm) | RMSE (mm) |
---|
Castle | 16 | 900 | 3600 | 18 | 82 | ~5 | 1.35 | 136 | 1.25 | 3.2 |
Rocca | 13 | 1000 | 14,000 | 22 | 93 | ~7 | 1.05 | 116 | 2.69 | 4.65 |
Table 4.
ZEB acquisition specifications.
Table 4.
ZEB acquisition specifications.
Specifications | Rocca | Castle |
---|
N° of scans | 11 | 2 |
Time | ~3 h | ~30 min |
N° of points | ~181,100,000 | ~35,410,000 |
Table 5.
Results of (I.) Dimensional and C2C comparison of ZEB and LiDAR point clouds on the Throne Chamber.
Table 5.
Results of (I.) Dimensional and C2C comparison of ZEB and LiDAR point clouds on the Throne Chamber.
Cases | n° of pts | | Dimensional Comparison (m) |  |
Length | Width | Height | Area 1 | Area 2 |
ZEB | 2,209,937 | | 12.6289 | 6.130 | 4.2117 | 75.9021 | 22.7629 |
TLS | 16,661,778 | | 12.6163 | 6.1231 | 4.2088 | 75.9006 | 22.7668 |
| | Δ | 0.0126 | 0.0069 | 0.0029 | 0.0015 | 0.0038 |
C2C comparison | No filter (Figure 11a) | Noise filter (Figure 11b) |
Mean | 0.0131 | 0.0076 |
St. dev. | 0.0214 | 0.0058 |
Table 6.
Results of C2C comparison ZEB point cloud and UAV Digital Surface Model (UAV DSM) on the Rocca volumes.
Table 6.
Results of C2C comparison ZEB point cloud and UAV Digital Surface Model (UAV DSM) on the Rocca volumes.
Cases | A | B | C | D |
---|
Complete ZEB-Complete UAV | Outdoor ZEB-Complete UAV | Outdoor ZEB-Optimized UAV | n°12 Points-Based Alignment |
---|
RMSE (cm) | 76.4 | 52.2 | 5.27 | 6.26 |
Min 1.33 |
Max 9.27 |
Table 7.
Comparison of point density related to the close-range photogrammetric model, LiDAR scan, and ZEB scan of the Fleur-de-lis Chamber. Similarities between (*) and between (#).
Table 7.
Comparison of point density related to the close-range photogrammetric model, LiDAR scan, and ZEB scan of the Fleur-de-lis Chamber. Similarities between (*) and between (#).
| (a) Close-Range SfM | (b) LiDAR | (c) ZEB |
---|
Original | Filtering |
---|
3 mm | 5 mm | 10 mm | 20 mm |
---|
N° points | 69,650,000 (*1) | 213,270,000 | 67,420,000 (*2) | 25,550,000 | 3,617,133 | 1,688,825 (#1) | 1,440,000 (#2) |
Density (pt/m2) | 197,800 | 556,000 | 185,000 | 60,000 | 15,000 | 3000 | 2400 |
Table 8.
The precision of coordinates derived from the test of the manual pick-point selection, statistically repeated five times on 20 points and applied on TLS and ZEB point cloud.
Table 8.
The precision of coordinates derived from the test of the manual pick-point selection, statistically repeated five times on 20 points and applied on TLS and ZEB point cloud.
5-Time Pick-Point | TLS-Extracted Coordinates | ZEB-Extracted Coordinates |
---|
(m) | X | Y | Z | Total | X | Y | Z | Total |
σmin | 0.0011 | 0.0024 | 0.0028 | 0.0021 | 0.0054 | 0.0044 | 0.0060 | 0.0053 |
σMAX | 0.0120 | 0.0101 | 0.0147 | 0.0123 | 0.0599 | 0.0287 | 0.0942 | 0.0609 |
σmean | 0.0047 | 0.0049 | 0.0061 | 0.0053 | 0.0163 | 0.0136 | 0.0249 | 0.0182 |
Table 9.
Results of the photogrammetric block adjustment using differently accurate set points: the topographic reference measurements on target points, the Ground Control Points (GCPs) extracted from the LiDAR point surface and the ones from the ZEB point cloud.
Table 9.
Results of the photogrammetric block adjustment using differently accurate set points: the topographic reference measurements on target points, the Ground Control Points (GCPs) extracted from the LiDAR point surface and the ones from the ZEB point cloud.
BBA RMSE | Topographic Coordinates | TLS-Extracted Coordinates | MMS-Extracted Coordinates |
---|
15 GCPs error (cm) | 0.43 | 0.95 | 7.72 |
5 CPs error (cm) | 0.54 | 1.37 | 11.03 |
Table 10.
Results of mesh triangulation process, based on different sensors point clouds.
Table 10.
Results of mesh triangulation process, based on different sensors point clouds.
Mesh Model | Close-Range SfM | LiDAR (5 mm) | ZEB | Fusion-Based |
---|
N° triangles | 13,937,030 | 5,110,790 | 318,232 | 3,391,239 |
*.obj (*.obj+ *.jpg) file size (Mb) | 1366 (1798+4) | 215 (386+9) | 24 (33+8) | 186 (297+9) |