|
|
Foundations of Robotics Seminar, April 25, 2007
Time and Place | Seminar
Abstract | Speaker Appointments
Surface Patch Reconstruction and Recognition via Curve
Sampling
Yan-Bin Jia
NSH 1507
Refreshments 4:15 pm
Talk 4:30 pm
We introduce a method that reconstructs a surface
patch by sampling along three concurrent curves on the surface with a touch
sensor. These data curves, each lying in
a different plane, form a "skeleton" from which the patch is built in
two phases. First, the Darboux frame at the curve intersection point is estimated
to reflect the local geometry. Second,
polynomial fitting is carried out under this frame. Simultaneous minimization of the total
(absolute) Gaussian curvature effectively prevents unnecessary foldings otherwise expected to result from the use of 1D
data. Experiments have demonstrated
remarkable accuracies of reconstruction.
Tactile data along concurrent curves can also be used
in model-based surface recognition. The
estimated Gaussian and mean curvatures at their intersection point are used in
a table lookup to find multiple candidate points on a surface model that have
similar local geometry. Starting at each
found point, local optimization is conducted to superpose the data onto the
model. Efficiency is achieved since the
data frame centered at the curve intersection has only three degrees of freedom
on the model. The best superposition is
chosen for comparison against other models for the recognition purpose.
For
appointments, please contact Yan-Bin Jia (jia at
cs.iastate.edu)
The Robotics
Institute is part of the School of
Computer Science, Carnegie Mellon University.