Unearthed Sydney 201726 Aug - 30 Aug
From Idea to Prototype on Resource Challenges in just a Weekend
This competition has finished.
Come rock with us, join the community
program you join.
resources industry. All are welcome.
Count the bolts - Geolocate rock bolts in an underground ore drive from 2D and 3D image data
Count the bolts aka Geolocate rock bolts in an underground ore drive from 2D and 3D image data.
•automatically locate bolts from 2D photos and 3D textured mesh
•provide georeferenced locations (CSV and visually)
•provide bolt statistics over a 20m section: total number, average distance apart, density per square meter
•classify condition of bolts: tight against wall or sticking out (stretch goal as imagery may not provide enough detail)
Rock bolts and mesh (ground support) are used to stop underground tunnels from collapsing inwards.
Currently a person will make a visual assessment on the pattern of bolting, as well as the physical condition of each bolt, to determine if ground support is holding well. Due to the sheer number of bolts the time taken to count and map them all is not practical.
A quick, automatic way to count and locate the bolts would make this assessment easier and more objective leading to an improved understanding of ground support performance and hence safer productive mining.
Georeferencing: In a GPS denied environment post processing of data occurs to align data to a particular coordinate system to enable overlay and analysis with GPS located information i.e. drill holes and survey.
*Please note the coordinates of this dataset have been shifted for client anonymity
Video on how bolting and meshing occurs https://youtu.be/CJ80Gbmes88
A blog about rock bolting
An MVS Sensor (LiDAR + Stereo Camera) was mounted to a 4WD light vehicle and driven along a section of underground tunnel. The following dataset is derived from a single scan of this section.
MVS Scan 20170217_140253
PGM: raw images
These are the raw images from the stereo camera.
ASC: reconstructed LiDAR point cloud
Raw LiDAR data has been processed into a 3D point cloud of the tunnel.
OBJ: Section B, 10cm textured mesh
A 3D mesh has been created from the LiDAR point cloud. Raw stereo imagery(texture) has been mapped to each 10cm triangle.
•bolt locations (x,y,z) as either CSV or visually in 2D or 3D images
•statistics as either CSV or data visualisation
•simple desktop app to process scan data and visualise results