fpfh.hpp

Go to the documentation of this file.

/*
 * Software License Agreement (BSD License)
 *
 *  Point Cloud Library (PCL) - www.pointclouds.org
 *  Copyright (c) 2010-2011, Willow Garage, Inc.
 *
 *  All rights reserved.
 *
 *  Redistribution and use in source and binary forms, with or without
 *  modification, are permitted provided that the following conditions
 *  are met:
 *
 *   * Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *   * Redistributions in binary form must reproduce the above
 *     copyright notice, this list of conditions and the following
 *     disclaimer in the documentation and/or other materials provided
 *     with the distribution.
 *   * Neither the name of Willow Garage, Inc. nor the names of its
 *     contributors may be used to endorse or promote products derived
 *     from this software without specific prior written permission.
 *
 *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
 *  FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
 *  COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
 *  INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 *  BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 *  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
 *  CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 *  LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
 *  ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 *  POSSIBILITY OF SUCH DAMAGE.
 *
 * $Id: fpfh.hpp 4927 2012-03-07 03:35:53Z rusu $
 *
 */

#ifndef PCL_FEATURES_IMPL_FPFH_H_
#define PCL_FEATURES_IMPL_FPFH_H_

#include <pcl/features/fpfh.h>
#include <pcl/features/pfh.h>

template <typename PointInT, typename PointNT, typename PointOutT> bool
pcl::FPFHEstimation<PointInT, PointNT, PointOutT>::computePairFeatures (
    const pcl::PointCloud<PointInT> &cloud, const pcl::PointCloud<PointNT> &normals,
    int p_idx, int q_idx, float &f1, float &f2, float &f3, float &f4)
{
  pcl::computePairFeatures (cloud.points[p_idx].getVector4fMap (), normals.points[p_idx].getNormalVector4fMap (),
      cloud.points[q_idx].getVector4fMap (), normals.points[q_idx].getNormalVector4fMap (),
      f1, f2, f3, f4);
  return (true);
}

template <typename PointInT, typename PointNT, typename PointOutT> void 
pcl::FPFHEstimation<PointInT, PointNT, PointOutT>::computePointSPFHSignature (
    const pcl::PointCloud<PointInT> &cloud, const pcl::PointCloud<PointNT> &normals,
    int p_idx, int row, const std::vector<int> &indices,
    Eigen::MatrixXf &hist_f1, Eigen::MatrixXf &hist_f2, Eigen::MatrixXf &hist_f3)
{
  Eigen::Vector4f pfh_tuple;
  // Get the number of bins from the histograms size
  int nr_bins_f1 = static_cast<int> (hist_f1.cols ());
  int nr_bins_f2 = static_cast<int> (hist_f2.cols ());
  int nr_bins_f3 = static_cast<int> (hist_f3.cols ());

  // Factorization constant
  float hist_incr = 100.0f / static_cast<float>(indices.size () - 1);

  // Iterate over all the points in the neighborhood
  for (size_t idx = 0; idx < indices.size (); ++idx)
  {
    // Avoid unnecessary returns
    if (p_idx == indices[idx])
        continue;

    // Compute the pair P to NNi
    if (!computePairFeatures (cloud, normals, p_idx, indices[idx], pfh_tuple[0], pfh_tuple[1], pfh_tuple[2], pfh_tuple[3]))
        continue;

    // Normalize the f1, f2, f3 features and push them in the histogram
    int h_index = static_cast<int> (floor (nr_bins_f1 * ((pfh_tuple[0] + M_PI) * d_pi_)));
    if (h_index < 0)           h_index = 0;
    if (h_index >= nr_bins_f1) h_index = nr_bins_f1 - 1;
    hist_f1 (row, h_index) += hist_incr;

    h_index = static_cast<int> (floor (nr_bins_f2 * ((pfh_tuple[1] + 1.0) * 0.5)));
    if (h_index < 0)           h_index = 0;
    if (h_index >= nr_bins_f2) h_index = nr_bins_f2 - 1;
    hist_f2 (row, h_index) += hist_incr;

    h_index = static_cast<int> (floor (nr_bins_f3 * ((pfh_tuple[2] + 1.0) * 0.5)));
    if (h_index < 0)           h_index = 0;
    if (h_index >= nr_bins_f3) h_index = nr_bins_f3 - 1;
    hist_f3 (row, h_index) += hist_incr;
  }
}

template <typename PointInT, typename PointNT, typename PointOutT> void
pcl::FPFHEstimation<PointInT, PointNT, PointOutT>::weightPointSPFHSignature (
    const Eigen::MatrixXf &hist_f1, const Eigen::MatrixXf &hist_f2, const Eigen::MatrixXf &hist_f3,
    const std::vector<int> &indices, const std::vector<float> &dists, Eigen::VectorXf &fpfh_histogram)
{
  assert (indices.size () == dists.size ());
  double sum_f1 = 0.0, sum_f2 = 0.0, sum_f3 = 0.0;
  float weight = 0.0, val_f1, val_f2, val_f3;

  // Get the number of bins from the histograms size
  int nr_bins_f1 = static_cast<int> (hist_f1.cols ());
  int nr_bins_f2 = static_cast<int> (hist_f2.cols ());
  int nr_bins_f3 = static_cast<int> (hist_f3.cols ());
  int nr_bins_f12 = nr_bins_f1 + nr_bins_f2;

  // Clear the histogram
  fpfh_histogram.setZero (nr_bins_f1 + nr_bins_f2 + nr_bins_f3);

  // Use the entire patch
  for (size_t idx = 0, data_size = indices.size (); idx < data_size; ++idx)
  {
    // Minus the query point itself
    if (dists[idx] == 0)
      continue;

    // Standard weighting function used
    weight = 1.0f / dists[idx];

    // Weight the SPFH of the query point with the SPFH of its neighbors
    for (int f1_i = 0; f1_i < nr_bins_f1; ++f1_i)
    {
      val_f1 = hist_f1 (indices[idx], f1_i) * weight;
      sum_f1 += val_f1;
      fpfh_histogram[f1_i] += val_f1;
    }

    for (int f2_i = 0; f2_i < nr_bins_f2; ++f2_i)
    {
      val_f2 = hist_f2 (indices[idx], f2_i) * weight;
      sum_f2 += val_f2;
      fpfh_histogram[f2_i + nr_bins_f1] += val_f2;
    }

    for (int f3_i = 0; f3_i < nr_bins_f3; ++f3_i)
    {
      val_f3 = hist_f3 (indices[idx], f3_i) * weight;
      sum_f3 += val_f3;
      fpfh_histogram[f3_i + nr_bins_f12] += val_f3;
    }
  }

  if (sum_f1 != 0)
    sum_f1 = 100.0 / sum_f1;           // histogram values sum up to 100
  if (sum_f2 != 0)
    sum_f2 = 100.0 / sum_f2;           // histogram values sum up to 100
  if (sum_f3 != 0)
    sum_f3 = 100.0 / sum_f3;           // histogram values sum up to 100

  // Adjust final FPFH values
  for (int f1_i = 0; f1_i < nr_bins_f1; ++f1_i)
    fpfh_histogram[f1_i] *= static_cast<float> (sum_f1);
  for (int f2_i = 0; f2_i < nr_bins_f2; ++f2_i)
    fpfh_histogram[f2_i + nr_bins_f1] *= static_cast<float> (sum_f2);
  for (int f3_i = 0; f3_i < nr_bins_f3; ++f3_i)
    fpfh_histogram[f3_i + nr_bins_f12] *= static_cast<float> (sum_f3);
}

template <typename PointInT, typename PointNT, typename PointOutT> void
pcl::FPFHEstimation<PointInT, PointNT, PointOutT>::computeSPFHSignatures (std::vector<int> &spfh_hist_lookup,
    Eigen::MatrixXf &hist_f1, Eigen::MatrixXf &hist_f2, Eigen::MatrixXf &hist_f3)
{
  // Allocate enough space to hold the NN search results
  // \note This resize is irrelevant for a radiusSearch ().
  std::vector<int> nn_indices (k_);
  std::vector<float> nn_dists (k_);

  std::set<int> spfh_indices;
  spfh_hist_lookup.resize (surface_->points.size ());

  // Build a list of (unique) indices for which we will need to compute SPFH signatures
  // (We need an SPFH signature for every point that is a neighbor of any point in input_[indices_])
  if (surface_ != input_ ||
      indices_->size () != surface_->points.size ())
  { 
    for (size_t idx = 0; idx < indices_->size (); ++idx)
    {
      int p_idx = (*indices_)[idx];
      if (this->searchForNeighbors (p_idx, search_parameter_, nn_indices, nn_dists) == 0)
        continue;

      spfh_indices.insert (nn_indices.begin (), nn_indices.end ());
    }
  }
  else
  {
    // Special case: When a feature must be computed at every point, there is no need for a neighborhood search
    for (size_t idx = 0; idx < indices_->size (); ++idx)
      spfh_indices.insert (static_cast<int> (idx));
  }

  // Initialize the arrays that will store the SPFH signatures
  size_t data_size = spfh_indices.size ();
  hist_f1.setZero (data_size, nr_bins_f1_);
  hist_f2.setZero (data_size, nr_bins_f2_);
  hist_f3.setZero (data_size, nr_bins_f3_);

  // Compute SPFH signatures for every point that needs them
  std::set<int>::iterator spfh_indices_itr = spfh_indices.begin ();
  for (int i = 0; i < static_cast<int> (spfh_indices.size ()); ++i)
  {
    // Get the next point index
    int p_idx = *spfh_indices_itr;
    ++spfh_indices_itr;

    // Find the neighborhood around p_idx
    if (this->searchForNeighbors (*surface_, p_idx, search_parameter_, nn_indices, nn_dists) == 0)
      continue;

    // Estimate the SPFH signature around p_idx
    computePointSPFHSignature (*surface_, *normals_, p_idx, i, nn_indices, hist_f1, hist_f2, hist_f3);

    // Populate a lookup table for converting a point index to its corresponding row in the spfh_hist_* matrices
    spfh_hist_lookup[p_idx] = i;
  }
}

template <typename PointInT, typename PointNT, typename PointOutT> void
pcl::FPFHEstimation<PointInT, PointNT, PointOutT>::computeFeature (PointCloudOut &output)
{
  // Allocate enough space to hold the NN search results
  // \note This resize is irrelevant for a radiusSearch ().
  std::vector<int> nn_indices (k_);
  std::vector<float> nn_dists (k_);

  std::vector<int> spfh_hist_lookup;
  computeSPFHSignatures (spfh_hist_lookup, hist_f1_, hist_f2_, hist_f3_);

  output.is_dense = true;
  // Save a few cycles by not checking every point for NaN/Inf values if the cloud is set to dense
  if (input_->is_dense)
  {
    // Iterate over the entire index vector
    for (size_t idx = 0; idx < indices_->size (); ++idx)
    {
      if (this->searchForNeighbors ((*indices_)[idx], search_parameter_, nn_indices, nn_dists) == 0)
      {
        for (int d = 0; d < fpfh_histogram_.size (); ++d)
          output.points[idx].histogram[d] = std::numeric_limits<float>::quiet_NaN ();
    
        output.is_dense = false;
        continue;
      }

      // ... and remap the nn_indices values so that they represent row indices in the spfh_hist_* matrices 
      // instead of indices into surface_->points
      for (size_t i = 0; i < nn_indices.size (); ++i)
        nn_indices[i] = spfh_hist_lookup[nn_indices[i]];

      // Compute the FPFH signature (i.e. compute a weighted combination of local SPFH signatures) ...
      weightPointSPFHSignature (hist_f1_, hist_f2_, hist_f3_, nn_indices, nn_dists, fpfh_histogram_);

      // ...and copy it into the output cloud
      for (int d = 0; d < fpfh_histogram_.size (); ++d)
        output.points[idx].histogram[d] = fpfh_histogram_[d];
    }
  }
  else
  {
    // Iterate over the entire index vector
    for (size_t idx = 0; idx < indices_->size (); ++idx)
    {
      if (!isFinite ((*input_)[(*indices_)[idx]]) ||
          this->searchForNeighbors ((*indices_)[idx], search_parameter_, nn_indices, nn_dists) == 0)
      {
        for (int d = 0; d < fpfh_histogram_.size (); ++d)
          output.points[idx].histogram[d] = std::numeric_limits<float>::quiet_NaN ();
    
        output.is_dense = false;
        continue;
      }

      // ... and remap the nn_indices values so that they represent row indices in the spfh_hist_* matrices 
      // instead of indices into surface_->points
      for (size_t i = 0; i < nn_indices.size (); ++i)
        nn_indices[i] = spfh_hist_lookup[nn_indices[i]];

      // Compute the FPFH signature (i.e. compute a weighted combination of local SPFH signatures) ...
      weightPointSPFHSignature (hist_f1_, hist_f2_, hist_f3_, nn_indices, nn_dists, fpfh_histogram_);

      // ...and copy it into the output cloud
      for (int d = 0; d < fpfh_histogram_.size (); ++d)
        output.points[idx].histogram[d] = fpfh_histogram_[d];
    }
  }
}

template <typename PointInT, typename PointNT> void
pcl::FPFHEstimation<PointInT, PointNT, Eigen::MatrixXf>::computeFeatureEigen (pcl::PointCloud<Eigen::MatrixXf> &output)
{
  // Set up the output channels
  output.channels["fpfh"].name     = "fpfh";
  output.channels["fpfh"].offset   = 0;
  output.channels["fpfh"].size     = 4;
  output.channels["fpfh"].count    = 33;
  output.channels["fpfh"].datatype = sensor_msgs::PointField::FLOAT32;

  // Allocate enough space to hold the NN search results
  // \note This resize is irrelevant for a radiusSearch ().
  std::vector<int> nn_indices (k_);
  std::vector<float> nn_dists (k_);

  std::vector<int> spfh_hist_lookup;
  this->computeSPFHSignatures (spfh_hist_lookup, hist_f1_, hist_f2_, hist_f3_);

  // Intialize the array that will store the FPFH signature
  output.points.resize (indices_->size (), nr_bins_f1_ + nr_bins_f2_ + nr_bins_f3_);
  output.is_dense = true;
  // Save a few cycles by not checking every point for NaN/Inf values if the cloud is set to dense
  if (input_->is_dense)
  {
    // Iterate over the entire index vector
    for (size_t idx = 0; idx < indices_->size (); ++idx)
    {
      if (!isFinite ((*input_)[(*indices_)[idx]]) ||
          this->searchForNeighbors ((*indices_)[idx], search_parameter_, nn_indices, nn_dists) == 0)
      {
        output.points.row (idx).setConstant (std::numeric_limits<float>::quiet_NaN ());
        output.is_dense = false;
        continue;
      }

      // ... and remap the nn_indices values so that they represent row indices in the spfh_hist_* matrices 
      // instead of indices into surface_->points
      for (size_t i = 0; i < nn_indices.size (); ++i)
        nn_indices[i] = spfh_hist_lookup[nn_indices[i]];

      // Compute the FPFH signature (i.e. compute a weighted combination of local SPFH signatures) ...
      this->weightPointSPFHSignature (hist_f1_, hist_f2_, hist_f3_, nn_indices, nn_dists, fpfh_histogram_);
      output.points.row (idx) = fpfh_histogram_;
    }
  }
  else
  {
    // Iterate over the entire index vector
    for (size_t idx = 0; idx < indices_->size (); ++idx)
    {
      if (this->searchForNeighbors ((*indices_)[idx], search_parameter_, nn_indices, nn_dists) == 0)
      {
        output.points.row (idx).setConstant (std::numeric_limits<float>::quiet_NaN ());
        output.is_dense = false;
        continue;
      }

      // ... and remap the nn_indices values so that they represent row indices in the spfh_hist_* matrices 
      // instead of indices into surface_->points
      for (size_t i = 0; i < nn_indices.size (); ++i)
        nn_indices[i] = spfh_hist_lookup[nn_indices[i]];

      // Compute the FPFH signature (i.e. compute a weighted combination of local SPFH signatures) ...
      this->weightPointSPFHSignature (hist_f1_, hist_f2_, hist_f3_, nn_indices, nn_dists, fpfh_histogram_);
      output.points.row (idx) = fpfh_histogram_;
    }
  }
}


#define PCL_INSTANTIATE_FPFHEstimation(T,NT,OutT) template class PCL_EXPORTS pcl::FPFHEstimation<T,NT,OutT>;

#endif    // PCL_FEATURES_IMPL_FPFH_H_