Migrated to ITK main

The FPFH module has been ingested into the Insight Toolkit (ITK) main repository and is no longer maintained here.

• In-tree location: Modules/Registration/FPFH
• Ingest PR: InsightSoftwareConsortium/ITK#6360
• Merge commit: 06172c1ec4

Future development, bug fixes, and pull requests should target ITK main. This repository's history is preserved for git blame and archival reference.

Maintainer action: this repository is Archived (read-only). See the ITK ingestion strategy for context.

---

ITKFPFH

Overview

Module to calculate FPFH feature for a pointset. Sample Usage is shown below:

# normal_pointset is ITK Pointset which contains normal vector for each point
# pointset is ITK Pointset which contains the input points for which feature needs to be calculated

# normal_np is numpy array of shape [Nx3]
# fpfh_feature is numpy array of shape [33xN]
# 25 is the radius and 100 is the maximum number of neighbors

pointset = itk.PointSet[itk.F, 3].New()
normal_pointset = itk.PointSet[itk.F, 3].New()

normal_pointset.SetPoints(itk.vector_container_from_array(normal_np.flatten()))
fpfh = itk.Fpfh.PointFeature.MF3MF3.New()
fpfh.ComputeFPFHFeature(pointset, normal_pointset, 25, 100)
fpfh_feature = fpfh.GetFpfhFeature()
fpfh_feature = itk.array_from_vector_container(fpfh_feature)
fpfh_feature = np.reshape(fpfh_feature, [33, pointset.GetNumberOfPoints()])

One can obtain the normals using the following code:

def getnormals_pca(inputPoints):
    import vtk
    from vtk.util import numpy_support
    meshPoints = numpy_to_vtk_polydata(inputPoints)
    normals = vtk.vtkPCANormalEstimation()
    normals.SetSampleSize(30)
    normals.SetFlipNormals(True)
    #normals.SetNormalOrientationToPoint()
    normals.SetNormalOrientationToGraphTraversal()
    normals.SetInputData(meshPoints)
    normals.Update()
    as_numpy = numpy_support.vtk_to_numpy(normals.GetOutput().GetPointData().GetArray(0))
    return as_numpy