Topology Computation (Tool)
(PCITopologyComputation)
Overview
This interactive tool computes statistical topology estimates for point
clouds. Currently, only unstructured point clouds are supported. The
tool assumes that a superset of the correct topology graph has already
been computed (use PCCComputeTopology
to create such a superset).
Parameters
Target:
- string targetNode - Target node in
scene graph. Can be selected by clicking into the scene.
Settings:
- bool in2d - Performs all computations
in 2d (assuming a flatland scene of lines)
- bool readStdDevFromPC - reads the
noise characteristics (std. deviation of unbiased Gaussian noise) from
an attribute channel (not implemented by now).
- float32 constStdDev - the constant
standard deviation that should be used if not beeing read from a
channel.
- bool readSamplingDensityFromPC -
reads the local sampling density from an attribute channel (not
implemented by now).
- float32 constSamplingDensity - the
constant sampling density used if not beeing read from a channel (not
implemented).
- bool colorPoints - if true, each
operation writes a suitable visualization into the color channel of the
point cloud.
- card32 uncertaintyTechnique - technique
used to estimate uncertainty in topology estimate. Choices are a PCA
technique or variance of the local model.
- card32 localModelEstTechnique -
technique used to estimate a local surface model (the topology is a
2d-chain of local models obeying to some consistency constraints).
choices are a least-squares or filtered voting ("kernel") based
technique. (2nd choice recommended).
Rendering:
- bool showUncertainty - draw little
balls to visualize the (local) standard deviation.
Methods
The following methods can be called from the ui:
- void computeUncertainty() - compute
the local model variance and write result into channel
localModelUncertainty (format 1f).
- bool computeTopology() - compute
statistical point cloud topology estimate - computes all other
properties as well as intermediate results.
Status
In development (incomplete).