The Geospatial Engineering department at Michigan Tech owns some LiDAR scanning solutions, thereof which includes a FARO Focus 3D. I—alongside others in the Douglass Houghton Student Chapter of the National Society of Professional Surveyors—performed some scans of snow sculptures during Winter Carnival, an annual event on campus that includes snow sculpture competitions.
We performed scans of two snow sculptures, done by Michigan Tech chapters of Delta Zeta + Sigma Phi Epsilon, and Phi Kappa Tau. Both were completed using the FARO Focus 3D.
Delta Zeta + Sigma Phi Epsilon's snow sculpture featured a large elephant with its trunk resting on a circus ball (this year's Winter Carnival theme was "carnival", given it was the 100th anniversary of the festival).
Phi Kappa Tau's snow sculpture featured a landscape of various structures in the Keweenaw area, a snow replica of the Michigan Tech Husky statue, and others.
The elephant scan was completed in 5 setups capturing ~25 million points, and the Phi Kappa Tau scan was completed in 11 setups collecting ~124 million points.
I then brought the raw capture data into FARO SCENE for processing. We opted out of not setting up targets onsite because we predicted SCENE would be able to register the scans based on surrounding geometry being mostly buildings and structures. This would prove to be true and both scans had mean point errors of less than 0.17 inches. You can view the full registration reports below.
Delta Zeta + Sigma Phi Epsilon point cloud
Phi Kappa Tau point cloud
Once the scans were processed in SCENE, I exported a mesh and brought that into Blender for cleaning. We desired to 3D print the elephant statue, so I had to make the mesh water-tight. We did not collect enough scans on-site to completely cover the elephant; there were gaps along its back, ears, and front legs. To fix this, I used a combination of MeshLab and Blender's sculpting toolset to refine the mesh.
A similar process was used for the Phi Kappa Tau scan as well.
We also completed a project of scanning the Husky Statue located on Michigan Tech's campus. We used a DJI Phantom 4 Pro to collect 34 pictures of the statue from different angles and elevations.
I used RealityCapture to stitch the photos together and generate a point cloud and mesh. Because our images did not have enough overlap for the software to fully register each angle, I had to manually insert some control points into the photos. After doing so, RealityCapture was able to register a cluster of 25/34 images.
After some cleaning of the mesh data, I was able to load the model into Cura and slice it for 3D printing.