Using Poisson Reconstruction as a Surface Regression Tool

[ryan1991991]

My team has been working on a problem involving scattered data in a 3D space which we need to regress onto a gridded surface. We have measurement data which has noise and can not be fit by a parametric function. Numerous other regression/interpolation methods have failed. The screened Poisson surface reconstruction acts as a tool which can fit a surface to any arbitrary collection of data points. We believe that this might be a useful tool to fit our data even though it is not traditional point cloud data used to recreate 3D models. This data does not have surface normals associated with it. Would the code be able to be modified to run with these constraints?

Thanks! Ryan

[mkazhdan]

Hi Ryan,

I have just posted an updated code-base (version 12.00) which should provide the support you are looking for. The function PointInterpolant takes as its input a set of positions and target values, and computes the coefficients of the function that best fits the values, while simultaneously minimizing a (bi-Harmonic) smoothness constraint. The function AdaptiveTreeVisualization has been modified to take the parameter "--samples". This points to a file containing a list of points at which a function is to be evaluated. When invoked with this argument, the AdaptiveTreeVisualization will sample the function at the prescribed positions and output the positions and values to standard out.

An example of usage can be found in "USAGE EXAMPLES (WITH SAMPLE DATA)" -> "PointInterpolant / AdaptiveTreeVisualization".

-- Misha

[Leerw]

@mkazhdan hi, is there any other links to download the executable .exe? I have to say I cannot open that link on README from China.

[mkazhdan]

I take the .zip file at: http://www.cs.jhu.edu/~misha/Code/PoissonRecon/Version12.00/AdaptiveSolvers.x64.zip doesn't work?

[Leerw]

I take the .zip file at: http://www.cs.jhu.edu/~misha/Code/PoissonRecon/Version12.00/AdaptiveSolvers.x64.zip doesn't work?

Thanks, this works.