Drones4Discovery: High resolution field data
using drones & digital pipelines
Steven Micklethwaite
Acknowledgements: Sam Thiele, Rob Glasgow, Monash University
DRONE DISCOVERY PLATFORM Visualisation
“To sense our world in greater detail, with greater quality
over greater areas, and to develop the technology and data
analytics that can do so for the benefit of Australian
research, industry and education.”
• Advanced remote sensing & sampling Data Analytics
dolerite
technologies + HPC/Cloud + analytics
tonalite
Capability
Developers
• For Research-Education-Industry Impact gabbro
• Driving interdisciplinary developments
Hyperspec
Capability
Users (Science-Engineering-IT-Arts)
• Regulatory oversight with CASA, including
.
pilots (protecting research)
• Leveraging MTRPs LiDAR
Flagship Workhorse
LiDAR Hyperspec (VNIR-SWIR) UAV-mag
6.1 X 104 nT
5.8 X 104 nT
0m 400 m
Visualisation
Thermal MEMS gravimetry
DATA ANALYTICS – CNN EXAMPLE
Ground truth labelled image
Manual digitisation: 51 mins
Background Vegetation
Water Dolerite
Original image (4000 x 6000 pixels) Aplite Tonalite
HSV input, random sampling (500), scale
variation
Time taken: 82.34 seconds
Pixel accuracy: 75.05 %
Dacite Gabbro
DATA FLUENCY: Photogrammetry example
EXTENDING PHASE
RAW DATA PHASE … FIELD
BASE MAP
… INFORMATION PHASE
FIELD DATA DATA INTERROGATION ANALYSIS VIZ. &
OBSERVATIONS COLLECTION PROCESSING EXPLORATION INTERPRETATION REPORTING
e.g.
3D point cloud
Orthophoto
DEM
e.g.
Filters
Bemis, Micklethwaite et al., Scalar products
JSG 2014 Some ML/DL
… TIME (opportunity/financial cost)
EXAMPLE LOCATION
Century Mine, Queensland
DATA PROCESSING: Photogrammetry example
Courtesy M. Roach (UTAS)
Photogrammetry (SfM-MVS);
Metashape/Photoscan, Pix4D etc
Reprojection Interpolation
orthophoto DEM
Feature Selection Bundle Adjustment Geometry Infill
SIFT Algorithm Interior and exterior dense point cloud
Bemis, Micklethwaite et al., JSG 2014
Texture Mapping Mesh Generation
photo-realistic model 3D model
INTERROGATION & EXPLORATION
METHOD DEVELOPMENT
Rapid fault, fracture, lithology
contact mapping
• Least-cost path solvers.
• Geologist remains involved in
interpretation.
• Future – CNN or other automated
Thiele et al., Solid Earth, 2017
ANALYSIS & INTERPRETATION
15 kilometres of dyke margin mapped @ 5-10 cm spatial resolution (fidelity)
50,000 orientation and thickness measurements
CLOUD COMPUTE: drones to information rapidly
e.g. PreViz, CAVE2
e.g. Figshare
CONCLUSIONS
& QUESTIONS
