Figure 1.
Samples of salt storage facilities: (a) rectangular barn and (b) dome.
Figure 1.
Samples of salt storage facilities: (a) rectangular barn and (b) dome.
Figure 2.
The integrated SMART system used in this study with two configurations: (a) tripod-mounted and (b) roof-mounted.
Figure 2.
The integrated SMART system used in this study with two configurations: (a) tripod-mounted and (b) roof-mounted.
Figure 3.
Illustration of the involved coordinate systems in the SMART system.
Figure 3.
Illustration of the involved coordinate systems in the SMART system.
Figure 4.
A sample point cloud (colored by LiDAR unit ID) and image captured by the SMART system at a given scan.
Figure 4.
A sample point cloud (colored by LiDAR unit ID) and image captured by the SMART system at a given scan.
Figure 5.
Data collection procedure for the SMART system and acquired LiDAR scans: (a) seven to thirteen scans are acquired at a given station and (b) limited overlap area between two successive scans.
Figure 5.
Data collection procedure for the SMART system and acquired LiDAR scans: (a) seven to thirteen scans are acquired at a given station and (b) limited overlap area between two successive scans.
Figure 6.
Locations of the three salt dome facilities used in this study along with a sample aerial view of each facility (aerial photo adopted from Google Earth Image).
Figure 6.
Locations of the three salt dome facilities used in this study along with a sample aerial view of each facility (aerial photo adopted from Google Earth Image).
Figure 7.
Demonstration of physical features constituting the dome storage facilities: (a) roof, (b) roof stringers, and (c) entrance walls (the SMART system is tripod-mounted at the center, roof-mounted at the center, and tripod-mounted at the entrance in these three examples, respectively).
Figure 7.
Demonstration of physical features constituting the dome storage facilities: (a) roof, (b) roof stringers, and (c) entrance walls (the SMART system is tripod-mounted at the center, roof-mounted at the center, and tripod-mounted at the entrance in these three examples, respectively).
Figure 8.
Proposed data processing framework for registration of SMART scans acquired in dome facilities.
Figure 8.
Proposed data processing framework for registration of SMART scans acquired in dome facilities.
Figure 9.
Image-aided coarse registration results for two successive scans in the West Lafayette dataset: (a) identified conjugate points between captured images in these two scans and (b) comparison of the before/post coarse registration results through a perspective view (middle) and two vertical profiles of the facility entrance (orange dotted box on the left) and the facility roof (pink dotted box on the right).
Figure 9.
Image-aided coarse registration results for two successive scans in the West Lafayette dataset: (a) identified conjugate points between captured images in these two scans and (b) comparison of the before/post coarse registration results through a perspective view (middle) and two vertical profiles of the facility entrance (orange dotted box on the left) and the facility roof (pink dotted box on the right).
Figure 10.
Examples of two successive scans colored by height (for easier illustration, the ground has been removed).
Figure 10.
Examples of two successive scans colored by height (for easier illustration, the ground has been removed).
Figure 11.
Schematic workflow of the proposed roof feature extraction procedure.
Figure 11.
Schematic workflow of the proposed roof feature extraction procedure.
Figure 12.
A sample roof segmentation result for a LiDAR scan in the West Lafayette dataset.
Figure 12.
A sample roof segmentation result for a LiDAR scan in the West Lafayette dataset.
Figure 13.
Coordinate system definition and parametric model for the roof feature with the axisymmetric characteristic of the roof feature (left) together with a vertical profile through the roof (right).
Figure 13.
Coordinate system definition and parametric model for the roof feature with the axisymmetric characteristic of the roof feature (left) together with a vertical profile through the roof (right).
Figure 14.
Workflow for the extraction and matching of roof stringers in SMART scans.
Figure 14.
Workflow for the extraction and matching of roof stringers in SMART scans.
Figure 15.
Schematic diagram of point-to-quadratic surface distance (left) and its 2D projection (right).
Figure 15.
Schematic diagram of point-to-quadratic surface distance (left) and its 2D projection (right).
Figure 16.
A sample edge point detection result for a roof feature in one scan (left: top view, right: perspective view).
Figure 16.
A sample edge point detection result for a roof feature in one scan (left: top view, right: perspective view).
Figure 17.
An example of 2D line segment detection results: (a) projected edge points, (b) detected 2D line segments (colored by id), and (c) point and line-segment outliers.
Figure 17.
An example of 2D line segment detection results: (a) projected edge points, (b) detected 2D line segments (colored by id), and (c) point and line-segment outliers.
Figure 18.
Examples of removed outliers: (a) point outliers (left: before, right: after) and (b) line-segment outliers. Line segments are colored by id (left: before, right: after).
Figure 18.
Examples of removed outliers: (a) point outliers (left: before, right: after) and (b) line-segment outliers. Line segments are colored by id (left: before, right: after).
Figure 19.
Illustration for the stringer matching: (a) definition of the stringer azimuth and (b) the successive matching procedure for mitigating the error propagation.
Figure 19.
Illustration for the stringer matching: (a) definition of the stringer azimuth and (b) the successive matching procedure for mitigating the error propagation.
Figure 20.
Established conjugate stringers for the West Lafayette dataset: (a) conjugate stringers from 13 scans (colored by stringer ID) and (b) a zoomed-in region for one stringer (colored by scan ID) (edge points are colored with grey).
Figure 20.
Established conjugate stringers for the West Lafayette dataset: (a) conjugate stringers from 13 scans (colored by stringer ID) and (b) a zoomed-in region for one stringer (colored by scan ID) (edge points are colored with grey).
Figure 21.
A review of the procedures involved in generating conjugate stringers for the West Lafayette dataset (all sub-images are shown in top view).
Figure 21.
A review of the procedures involved in generating conjugate stringers for the West Lafayette dataset (all sub-images are shown in top view).
Figure 22.
Demonstration of planar features near the entrance of a salt dome facility: (a) tripod-mounted SMART system stationed near the entrance and (b) wall/ground features.
Figure 22.
Demonstration of planar features near the entrance of a salt dome facility: (a) tripod-mounted SMART system stationed near the entrance and (b) wall/ground features.
Figure 23.
A LiDAR scan and two sample seed points; points are represented by Cartesian (left) and normalized polar (right) coordinates.
Figure 23.
A LiDAR scan and two sample seed points; points are represented by Cartesian (left) and normalized polar (right) coordinates.
Figure 24.
Planar feature segmentation results: (a) comparison of segmentation results based on Cartesian and normalized polar coordinates of the original point cloud (ground and entrance walls are highlighted in blue and red dotted boxes, respectively), (b) zoom-in window of the ground feature, and (c) zoom-in window of the entrance wall feature.
Figure 24.
Planar feature segmentation results: (a) comparison of segmentation results based on Cartesian and normalized polar coordinates of the original point cloud (ground and entrance walls are highlighted in blue and red dotted boxes, respectively), (b) zoom-in window of the ground feature, and (c) zoom-in window of the entrance wall feature.
Figure 25.
Illustration of the stringer-specific coordinate system.
Figure 25.
Illustration of the stringer-specific coordinate system.
Figure 26.
Three forms for representing a plane based on the orientation of the normal vector.
Figure 26.
Three forms for representing a plane based on the orientation of the normal vector.
Figure 27.
Perspective view (left) and top view (right) of the extracted roof feature for: (a) Lebanon, (b) Frankfort, and (c) West Lafayette datasets (points are colored by scan ID with the red circle showing the outline of the facility).
Figure 27.
Perspective view (left) and top view (right) of the extracted roof feature for: (a) Lebanon, (b) Frankfort, and (c) West Lafayette datasets (points are colored by scan ID with the red circle showing the outline of the facility).
Figure 28.
Extracted vertical profiles (width: 0.2 m) for the Lebanon dataset before/after partial fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 28.
Extracted vertical profiles (width: 0.2 m) for the Lebanon dataset before/after partial fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 29.
Extracted vertical profiles (width: 0.2 m) from the Frankfort dataset before/after partial fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 29.
Extracted vertical profiles (width: 0.2 m) from the Frankfort dataset before/after partial fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 30.
Extracted vertical profiles (width: 0.2 m) from the West Lafayette dataset before/after partial fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 30.
Extracted vertical profiles (width: 0.2 m) from the West Lafayette dataset before/after partial fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 31.
Heat maps and box plots representing the normal distance between the roof feature points and fitted quadratic surface for the Lebanon dataset before (left) and after (right) partial fine registration.
Figure 31.
Heat maps and box plots representing the normal distance between the roof feature points and fitted quadratic surface for the Lebanon dataset before (left) and after (right) partial fine registration.
Figure 32.
Heat maps and box plots representing the normal distance between the roof feature points and fitted quadratic surface for the Frankfort dataset before (left) and after (right) partial fine registration.
Figure 32.
Heat maps and box plots representing the normal distance between the roof feature points and fitted quadratic surface for the Frankfort dataset before (left) and after (right) partial fine registration.
Figure 33.
Heat maps and box plots representing the normal distance between the roof feature points and fitted quadratic surface for the West Lafayette dataset before (left) and after (right) partial fine registration.
Figure 33.
Heat maps and box plots representing the normal distance between the roof feature points and fitted quadratic surface for the West Lafayette dataset before (left) and after (right) partial fine registration.
Figure 34.
Extracted vertical profiles (width: 0.2 m) through the salt surface of the Lebanon dataset before/after full fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 34.
Extracted vertical profiles (width: 0.2 m) through the salt surface of the Lebanon dataset before/after full fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 35.
Extracted edge points from all scans of the Lebanon dataset (a) before and (b) after full fine registration.
Figure 35.
Extracted edge points from all scans of the Lebanon dataset (a) before and (b) after full fine registration.
Figure 36.
Extracted stringer features and the corresponding feature IDs for the Lebanon dataset.
Figure 36.
Extracted stringer features and the corresponding feature IDs for the Lebanon dataset.
Figure 37.
Extracted vertical profiles (width: 0.2 m) from the salt surface of the Frankfort dataset after the partial fine registration using the roof feature, partial fine registration using the ground/wall features, and full fine registration using the roof/ground/wall features (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 37.
Extracted vertical profiles (width: 0.2 m) from the salt surface of the Frankfort dataset after the partial fine registration using the roof feature, partial fine registration using the ground/wall features, and full fine registration using the roof/ground/wall features (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 38.
Extracted edge points from the scans in the Frankfort dataset showing the tangential alignment of the radial stringers after the: (a) partial fine registration using the roof feature, (b) partial fine registration using the ground/wall features, and (c) full fine registration using the roof/ground/wall features.
Figure 38.
Extracted edge points from the scans in the Frankfort dataset showing the tangential alignment of the radial stringers after the: (a) partial fine registration using the roof feature, (b) partial fine registration using the ground/wall features, and (c) full fine registration using the roof/ground/wall features.
Figure 39.
Extracted edge points from the scans in the Frankfort dataset showing the radial alignment of the stringer rings after the: (a) partial fine registration using the roof feature, (b) partial fine registration using the ground/wall features, and (c) full fine registration using the roof/ground/wall features.
Figure 39.
Extracted edge points from the scans in the Frankfort dataset showing the radial alignment of the stringer rings after the: (a) partial fine registration using the roof feature, (b) partial fine registration using the ground/wall features, and (c) full fine registration using the roof/ground/wall features.
Figure 40.
Extracted vertical profiles (width: 0.2 m) along the salt surface of the West Lafayette dataset before/after the full fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 40.
Extracted vertical profiles (width: 0.2 m) along the salt surface of the West Lafayette dataset before/after the full fine registration (the arrows indicate the direction of the measured distance): (a) profile orthogonal to the entrance walls and (b) profile parallel to the entrance walls.
Figure 41.
Extracted edge points from the scans in the West Lafayette dataset before/after the full fine registration: (a) partial fine registered edge points, (b) full fine registered edge points using roof/stringer features, and (c) full fine registered edge points using the roof/stringer/ground/wall features.
Figure 41.
Extracted edge points from the scans in the West Lafayette dataset before/after the full fine registration: (a) partial fine registered edge points, (b) full fine registered edge points using roof/stringer features, and (c) full fine registered edge points using the roof/stringer/ground/wall features.
Figure 42.
Extracted stringer features and the corresponding feature IDs for the West Lafayette dataset.
Figure 42.
Extracted stringer features and the corresponding feature IDs for the West Lafayette dataset.
Table 1.
Data collection sites, facility size, number of scans, and system location.
Table 1.
Data collection sites, facility size, number of scans, and system location.
Salt Dome Facility | Size (m) | Number of Scans | System Location |
---|
Radius | Height |
---|
Lebanon unit | 10.0 | 11.0 | 9 | Inside the facility |
Frankfort unit | 14.5 | 17 | 12 | By the entrance |
West Lafayette unit | 13.5 | 15.5 | 13 | Inside the facility |
Table 2.
Availability of registration features(shown in red) in the Lebanon, Frankfort, and West Lafayette datasets.
Table 2.
Availability of registration features(shown in red) in the Lebanon, Frankfort, and West Lafayette datasets.
Salt Dome Facility | Lebanon | Frankfort | West Lafayette |
---|
SMART System Information | Mount | Location | Mount | Location | Mount | Location |
---|
Roof | Center | Tripod | Entrance | Tripod | Center |
---|
Feature availability | Roof | Visible | Visible | Visible |
Stringers | Visible | Not clearly visible | Visible |
Ground | Not visible | Visible | Visible in most scans |
Walls | Not visible | Visible | Visible in most scans |
Table 3.
RMS of the fitting error for the roof and ground features following the coarse/partial fine registration.
Table 3.
RMS of the fitting error for the roof and ground features following the coarse/partial fine registration.
Roof Quality of Fit (Utilized in LSA) |
Dataset | Number of Points | RMS of Normal Distance (m) |
Coarse | Partial |
Lebanon | 339,648 | 0.046 | 0.044 |
Frankfort | 276,769 | 0.095 | 0.051 |
West Lafayette | 415,896 | 0.085 | 0.067 |
Ground Quality of Fit (not Utilized in LSA) |
Dataset | Number of Points | RMS of Normal Distance (m) |
Coarse | Partial |
Lebanon | - | - | - |
Frankfort | 181,846 | 0.229 | 0.071 |
West Lafayette | 115,867 | 0.154 | 0.053 |
Table 4.
RMS of the fitting error for the registration primitives (roof and stringers) in the Lebanon dataset following the partial/full fine registration.
Table 4.
RMS of the fitting error for the registration primitives (roof and stringers) in the Lebanon dataset following the partial/full fine registration.
Roof Quality of Fit |
Number of points | RMS of Normal Distance (m) |
Partial | Full |
316,155 | 0.038 | 0.040 |
Stringer Quality of Fit |
Stringer ID | Number of points | RMS of Normal Distance (m) |
Partial | Full |
1 | 1505 | 0.117 | 0.105 |
2 | 1440 | 0.121 | 0.109 |
3 | 1286 | 0.122 | 0.110 |
4 | 1123 | 0.129 | 0.112 |
5 | 1096 | 0.130 | 0.114 |
6 | 1075 | 0.146 | 0.109 |
7 | 1030 | 0.113 | 0.081 |
8 | 867 | 0.098 | 0.070 |
9 | 682 | 0.093 | 0.069 |
10 | 852 | 0.115 | 0.089 |
11 | 487 | 0.124 | 0.112 |
12 | 1485 | 0.134 | 0.124 |
13 | 1464 | 0.121 | 0.113 |
14 | 1346 | 0.111 | 0.102 |
15 | 1478 | 0.106 | 0.098 |
16 | 1536 | 0.102 | 0.093 |
17 | 1322 | 0.092 | 0.082 |
18 | 1194 | 0.090 | 0.081 |
19 | 995 | 0.096 | 0.085 |
20 | 1245 | 0.108 | 0.096 |
Table 5.
RMS of the fitting error for the registration primitives in the Frankfort dataset after partial fine registration using the roof feature (partial 1), partial fine registration using the ground/wall (partial 2) features, and full fine registration using the roof/ground/wall features.
Table 5.
RMS of the fitting error for the registration primitives in the Frankfort dataset after partial fine registration using the roof feature (partial 1), partial fine registration using the ground/wall (partial 2) features, and full fine registration using the roof/ground/wall features.
Roof Quality of Fit |
Number of Points | RMS of Normal Distance (m) |
Partial 1 | Partial 2 | Full |
276,769 | 0.051 | 0.086 | 0.052 |
Ground Quality of Fit |
Number of Points | RMS of Normal Distance (m) |
Partial 1 | Partial 2 | Full |
181,846 | 0.071 | 0.033 | 0.033 |
Vertical Walls Quality of Fit |
Left | Right |
Number of Points | RMS of Normal Distance (m) | Number of Points | RMS of Normal Distance (m) |
Partial 1 | Partial 2 | Full | Partial 1 | Partial 2 | Full |
34,951 | 0.190 | 0.046 | 0.047 | 26,688 | 0.220 | 0.047 | 0.048 |
Table 6.
RMS of the fitting error of the registration primitives in the West Lafayette dataset after partial fine registration using the roof feature, full fine registration using the roof/stringer features (Full 1), and full fine registration using the roof/stringer/ground/wall features (Full 2).
Table 6.
RMS of the fitting error of the registration primitives in the West Lafayette dataset after partial fine registration using the roof feature, full fine registration using the roof/stringer features (Full 1), and full fine registration using the roof/stringer/ground/wall features (Full 2).
Roof Quality of Fit |
Number of Points | RMS of Normal Distance (m) |
Partial | Full 1 | Full 2 |
403,158 | 0.066 | 0.067 | 0.072 |
Stringer Quality of Fit |
StringerID | Number of Points | RMS of Normal Distance (m) |
Partial | Full 1 | Full 2 |
1 | 314 | 0.145 | 0.094 | 0.100 |
2 | 350 | 0.148 | 0.103 | 0.117 |
3 | 302 | 0.165 | 0.133 | 0.143 |
4 | 288 | 0.188 | 0.143 | 0.156 |
5 | 323 | 0.209 | 0.188 | 0.191 |
6 | 212 | 0.208 | 0.195 | 0.210 |
7 | 80 | 0.124 | 0.115 | 0.118 |
8 | 189 | 0.094 | 0.089 | 0.069 |
9 | 350 | 0.122 | 0.075 | 0.055 |
10 | 248 | 0.108 | 0.091 | 0.062 |
11 | 204 | 0.108 | 0.079 | 0.062 |
12 | 254 | 0.149 | 0.082 | 0.079 |
13 | 324 | 0.139 | 0.096 | 0.094 |
14 | 303 | 0.202 | 0.156 | 0.135 |
15 | 611 | 0.202 | 0.182 | 0.195 |
16 | 615 | 0.190 | 0.182 | 0.185 |
17 | 526 | 0.144 | 0.105 | 0.116 |
18 | 471 | 0.078 | 0.057 | 0.070 |
19 | 644 | 0.084 | 0.053 | 0.065 |
20 | 524 | 0.128 | 0.073 | 0.077 |
Ground Quality of Fit |
Number of Points | RMS of Normal Distance (m) |
Partial | Full 1 | Full 2 |
115,867 | 0.053 | 0.050 | 0.029 |
Vertical Walls Quality of Fit |
Left | Right |
Number of Points | RMS of Normal Distance (m) | Number of Points | RMS of Normal Distance (m) |
Partial | Full 1 | Full 2 | Partial | Full 1 | Full 2 |
- | - | - | - | 915 | 0.126 | 0.031 | 0.013 |