5
cloud data is too large, the 3D mesh was generated with a
low triangulation count which results in a smoother and
faster rendering scene, without losing accuracy when uti-
lizing interpolation over the surface of the mesh. The 3D
mesh is then exported as a .ply file from Cloud Compare
and is imported into Blender for scaling and isometric
embedding, which is also called UV unwrapping done
with the unwrapping tool. This method of generating 3D
meshes results in more realistic and spatially accurate 3D
models (Figure 2). However, due to the complexity of the
3D mesh, UV unwrapping can be challenging. The acquisi-
tion of point cloud data can be lengthy and requires person-
nel or machinery as mentioned in Section 3, convoluting
the process to update and develop 3D models as advance-
ments are made. Nevertheless, georeferenced point cloud
data can ensure the rendered visualizations are accurately
representative of the decline.
Another way the 3D mesh is generated is by creating
simplistic, representative 3D solid models of the openings
at the East decline that match real-world scale. This method
was done directly in Blender by utilizing the modeling tool
to manually create a solid object with a horseshoe arc shape
and the dimensions of 5-meters in height and 5-meters
in width (Figure 3). A representative model could also be
created by utilizing .dxf or .shp files that were developed
during the planning stages. The 3D model is then scaled,
and UV unwrapped. This process compared to generating
a 3D mesh from point cloud data requires more hands-on
work in Blender. Also, the model is a representation and
might not show precise location of the movement changes
in the decline as advancement takes place, and inaccuracies
should be kept in mind when visualizations are created for
more advanced numerical methods in geotechnical analy-
sis. However, since this process does not require 3D point
cloud data, scanning of the decline is not required which
can ensure worker safety and save time in the long run as
simplistic 3D models can be utilized whenever needed.
Therefore, an optimal methodological approach will be
to utilize both ways, where mesh generation in the short
term uses representative models and in the long term uses
updated complex models from scanning data.
After generating the 3D mesh, the triangulations need
to be represented in a 2D space to define each triangle in
each vertex of the 3D mesh UV coordinate. This process is
similar to “mapping” 3D objects into 2D spaces, such as
generating an accurate and reliable 2D map of the Earth,
which is a sphere-like 3D object. The UV coordinate system
is a 2D coordinate system similar to a cartesian coordinate
system. It is used to translate each vertex of the triangu-
lations that are in 3D space onto a 2D space, through
isometric embedding. The isometric embedding is a semi-
automated process using the unwrapping function of the
unwrapping tool in Blender. Even though this method
does not require an isometric embedding algorithm to be
developed, the 3D mesh needs to be laid out onto the UV
coordinate system through the manual efforts of a miner.
This process is crucial as it will determine the scale of the
2D representation of the 3D mesh which in return affects
the distance used in the interpolation of displacement data.
It also determines the horizontal and the vertical space as
U represents the X axis, horizontal space, and V represents
the Y axis, vertical space. This study unwraps the 3D mesh
into five major segments that consist of the face, the floor,
the back, and the left and right ribs of the underground
opening (Figure 4).
Differentiation of the Spaces on a 3D Mesh
This step in the methodology can be done in various ways
with various complexities using HLSL. Thus, for this study,
the decline at the SX is differentiated into 5 segments,
the left rib, the right rib, the back, the floor, and the face.
Figure 2. Complex and realistic 3D mesh of the East decline
generated using point cloud data
Figure 3. Representative and simple 3D mesh generated in
Blender according to profile shape and dimensions
cloud data is too large, the 3D mesh was generated with a
low triangulation count which results in a smoother and
faster rendering scene, without losing accuracy when uti-
lizing interpolation over the surface of the mesh. The 3D
mesh is then exported as a .ply file from Cloud Compare
and is imported into Blender for scaling and isometric
embedding, which is also called UV unwrapping done
with the unwrapping tool. This method of generating 3D
meshes results in more realistic and spatially accurate 3D
models (Figure 2). However, due to the complexity of the
3D mesh, UV unwrapping can be challenging. The acquisi-
tion of point cloud data can be lengthy and requires person-
nel or machinery as mentioned in Section 3, convoluting
the process to update and develop 3D models as advance-
ments are made. Nevertheless, georeferenced point cloud
data can ensure the rendered visualizations are accurately
representative of the decline.
Another way the 3D mesh is generated is by creating
simplistic, representative 3D solid models of the openings
at the East decline that match real-world scale. This method
was done directly in Blender by utilizing the modeling tool
to manually create a solid object with a horseshoe arc shape
and the dimensions of 5-meters in height and 5-meters
in width (Figure 3). A representative model could also be
created by utilizing .dxf or .shp files that were developed
during the planning stages. The 3D model is then scaled,
and UV unwrapped. This process compared to generating
a 3D mesh from point cloud data requires more hands-on
work in Blender. Also, the model is a representation and
might not show precise location of the movement changes
in the decline as advancement takes place, and inaccuracies
should be kept in mind when visualizations are created for
more advanced numerical methods in geotechnical analy-
sis. However, since this process does not require 3D point
cloud data, scanning of the decline is not required which
can ensure worker safety and save time in the long run as
simplistic 3D models can be utilized whenever needed.
Therefore, an optimal methodological approach will be
to utilize both ways, where mesh generation in the short
term uses representative models and in the long term uses
updated complex models from scanning data.
After generating the 3D mesh, the triangulations need
to be represented in a 2D space to define each triangle in
each vertex of the 3D mesh UV coordinate. This process is
similar to “mapping” 3D objects into 2D spaces, such as
generating an accurate and reliable 2D map of the Earth,
which is a sphere-like 3D object. The UV coordinate system
is a 2D coordinate system similar to a cartesian coordinate
system. It is used to translate each vertex of the triangu-
lations that are in 3D space onto a 2D space, through
isometric embedding. The isometric embedding is a semi-
automated process using the unwrapping function of the
unwrapping tool in Blender. Even though this method
does not require an isometric embedding algorithm to be
developed, the 3D mesh needs to be laid out onto the UV
coordinate system through the manual efforts of a miner.
This process is crucial as it will determine the scale of the
2D representation of the 3D mesh which in return affects
the distance used in the interpolation of displacement data.
It also determines the horizontal and the vertical space as
U represents the X axis, horizontal space, and V represents
the Y axis, vertical space. This study unwraps the 3D mesh
into five major segments that consist of the face, the floor,
the back, and the left and right ribs of the underground
opening (Figure 4).
Differentiation of the Spaces on a 3D Mesh
This step in the methodology can be done in various ways
with various complexities using HLSL. Thus, for this study,
the decline at the SX is differentiated into 5 segments,
the left rib, the right rib, the back, the floor, and the face.
Figure 2. Complex and realistic 3D mesh of the East decline
generated using point cloud data
Figure 3. Representative and simple 3D mesh generated in
Blender according to profile shape and dimensions