Team members: Alexei Masterov - David Smilowitz
- Alejandro Troccoli - Sam Neymotin
Range Data Acquisiton
Le Penseur is located on Columbia's main campus, to the west of Philosophy
hall and North of Kent Hall, as marked in yellow below.
Location of Le Penseur.
The scanning of Le Penseur took place on October 18th over the course of
3 hours. For registration purposes, we used a total of 16 targets. Three
targets were attached to each side of the base and 1 target was mounted to
each tower. There were 4 towers positioned around the base of the sculpture.
This gave us a total of 16 targets to use for registration. Our first 4
scans were planned in such a way that we could capture 8 targets on each.
This was done so that any additional scans only had to contain 3 targets.
After taking the first 4 scans we registered them in Cyclone and planed
the additional scans. Based on our observations we decided that 5 additional
scans were needed to fill in the holes. Three of these scans were taken
from the ground level and 2 of them from two windows on the 5 floor of Philosophy
Hall. The French Department was extremely helpful since they allowed us
to use their offices to scan the top of the sculpture.
Registration of the scans was done using the Cyclone software and the
targets. As mentioned above the first 4 scans were registered first. Then
we added the 2 scans from the windows and finally the 3 remaining ground
scans. This allowed us to detect errors in the target acquisition. We had
problems with two targets because in both cases a tower partly occluded a
target on the base. This partial occlusion caused the scanner to think the
target was located on the tower when in reality it was on the base of the
sculpture. We were able to identify this problem and removed the occluded
targets from the registration process. Once we removed the misplaced targets
the resulting registration was accurate.
The nine scans
Analysis of Scanning and Registration
Overall the scanning and registration process was a success. However there
are some improvements that could be made. The main problem turned out to
be the fact that the number of high scanning locations was limited. This
was due to the positioning of nearby buildings and the presence of trees.
The result was that only a minimal amount of data about some areas, such
as the right hand, was captured. This lack of data made the meshing process
much more difficult. The quality of the final point cloud would be increased
if additional scans were taken from a tall moveable platform. Of course
a more advanced scanner would also improve the detail contained in the model.