The Plane Fit Criterion was introduced in [Weyrich2004] and computes the deviation of a point from a manifold approximating its neighbors. First, a plane is fit to each point’s k-neighborhood by performing an eigenvalue decomposition. Next, the mean point to plane distance is computed by considering all points within the neighborhood. This is compared to the point to plane distance of the current point giving rise to the k-neighborhood. As the mean distance of the k-neighborhood approaches 0, the Plane Fit criterion will tend toward 1. As point to plane distance of the current point approaches 0, the Plane Fit criterion will tend toward 0.
The author suggests that the Plane Fit Criterion is well suited to outlier detection when considering noisy reconstructions of smooth surfaces, but produces poor results around small features and creases.
The filter creates a single new dimension,
PlaneFit, that records the
Plane Fit criterion for the current point.
To inspect the newly created, non-standard dimensions, be sure to write to an output format that can support arbitrary dimensions, such as BPF.
The sample pipeline below computes the Plane Fit criterion with a neighborhood of 8 neighbors. We do not apply a fixed threshold to single out outliers based on the Plane Fit criterion as the range of values can vary from one dataset to another. In general, higher values indicate the likelihood of a point being an outlier.
The number of k nearest neighbors. [Default: 8]
The number of threads used for computing the plane fit criterion. [Default: 1]
An expression that limits points passed to a filter. Points that don’t pass the expression skip the stage but are available to subsequent stages in a pipeline. [Default: no filtering]
A strategy for merging points skipped by a ‘where’ option when running in standard mode. If
true, the skipped points are added to the first point view returned by the skipped filter. If
false, skipped points are placed in their own point view. If
auto, skipped points are merged into the returned point view provided that only one point view is returned and it has the same point count as it did when the filter was run. [Default: