organized_multi_plane_segmentation.hpp

/*
 * Software License Agreement (BSD License)
 *
 *  Point Cloud Library (PCL) - www.pointclouds.org
 *  Copyright (c) 2010-2012, 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 the copyright holder(s) 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.
 *
 *
 *
 */

#ifndef PCL_SEGMENTATION_IMPL_ORGANIZED_MULTI_PLANE_SEGMENTATION_H_
#define PCL_SEGMENTATION_IMPL_ORGANIZED_MULTI_PLANE_SEGMENTATION_H_

#include <pcl/segmentation/boost.h>
#include <pcl/segmentation/organized_connected_component_segmentation.h>
#include <pcl/segmentation/organized_multi_plane_segmentation.h>
#include <pcl/common/centroid.h>
#include <pcl/common/eigen.h>

/////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template <typename PointT> pcl::PointCloud<PointT>
projectToPlaneFromViewpoint (pcl::PointCloud<PointT>& cloud, Eigen::Vector4f& normal, Eigen::Vector3f& centroid, Eigen::Vector3f& vp)
{
  Eigen::Vector3f norm (normal[0], normal[1], normal[2]); //(region.coefficients_[0], region.coefficients_[1], region.coefficients_[2]); 
  pcl::PointCloud<PointT> projected_cloud;
  projected_cloud.resize (cloud.points.size ());
  for (size_t i = 0; i < cloud.points.size (); i++)
  {
    Eigen::Vector3f pt (cloud.points[i].x, cloud.points[i].y, cloud.points[i].z);
    //Eigen::Vector3f intersection = (vp, pt, norm, centroid);
    float u = norm.dot ((centroid - vp)) / norm.dot ((pt - vp));
    Eigen::Vector3f intersection (vp + u * (pt - vp));
    projected_cloud[i].x = intersection[0];
    projected_cloud[i].y = intersection[1];
    projected_cloud[i].z = intersection[2];
  }
  
  return (projected_cloud);
}

//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template<typename PointT, typename PointNT, typename PointLT> void
pcl::OrganizedMultiPlaneSegmentation<PointT, PointNT, PointLT>::segment (std::vector<ModelCoefficients>& model_coefficients, 
                                                                         std::vector<PointIndices>& inlier_indices)
{
  pcl::PointCloud<pcl::Label> labels;
  std::vector<pcl::PointIndices> label_indices;
  std::vector<Eigen::Vector4f, Eigen::aligned_allocator<Eigen::Vector4f> > centroids;
  std::vector <Eigen::Matrix3f, Eigen::aligned_allocator<Eigen::Matrix3f> > covariances;
  segment (model_coefficients, inlier_indices, centroids, covariances, labels, label_indices);
}

//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template<typename PointT, typename PointNT, typename PointLT> void
pcl::OrganizedMultiPlaneSegmentation<PointT, PointNT, PointLT>::segment (std::vector<ModelCoefficients>& model_coefficients, 
                                                                         std::vector<PointIndices>& inlier_indices,
                                                                         std::vector<Eigen::Vector4f, Eigen::aligned_allocator<Eigen::Vector4f> >& centroids,
                                                                         std::vector <Eigen::Matrix3f, Eigen::aligned_allocator<Eigen::Matrix3f> >& covariances,
                                                                         pcl::PointCloud<PointLT>& labels,
                                                                         std::vector<pcl::PointIndices>& label_indices)
{
  if (!initCompute ())
    return;

  // Check that we got the same number of points and normals
  if (static_cast<int> (normals_->points.size ()) != static_cast<int> (input_->points.size ()))
  {
    PCL_ERROR ("[pcl::%s::segment] Number of points in input cloud (%lu) and normal cloud (%lu) do not match!\n",
               getClassName ().c_str (), input_->points.size (),
               normals_->points.size ());
    return;
  }

  // Check that the cloud is organized
  if (!input_->isOrganized ())
  {
    PCL_ERROR ("[pcl::%s::segment] Organized point cloud is required for this plane extraction method!\n",
               getClassName ().c_str ());
    return;
  }

  // Calculate range part of planes' hessian normal form
  std::vector<float> plane_d (input_->points.size ());
  
  for (unsigned int i = 0; i < input_->size (); ++i)
    plane_d[i] = input_->points[i].getVector3fMap ().dot (normals_->points[i].getNormalVector3fMap ());
  
  // Make a comparator
  //PlaneCoefficientComparator<PointT,PointNT> plane_comparator (plane_d);
  compare_->setPlaneCoeffD (plane_d);
  compare_->setInputCloud (input_);
  compare_->setInputNormals (normals_);
  compare_->setAngularThreshold (static_cast<float> (angular_threshold_));
  compare_->setDistanceThreshold (static_cast<float> (distance_threshold_), true);

  // Set up the output
  OrganizedConnectedComponentSegmentation<PointT,PointLT> connected_component (compare_);
  connected_component.setInputCloud (input_);
  connected_component.segment (labels, label_indices);

  Eigen::Vector4f clust_centroid = Eigen::Vector4f::Zero ();
  Eigen::Vector4f vp = Eigen::Vector4f::Zero ();
  Eigen::Matrix3f clust_cov;
  pcl::ModelCoefficients model;
  model.values.resize (4);

  // Fit Planes to each cluster
  for (size_t i = 0; i < label_indices.size (); i++)
  {
    if (static_cast<unsigned> (label_indices[i].indices.size ()) > min_inliers_)
    {
      pcl::computeMeanAndCovarianceMatrix (*input_, label_indices[i].indices, clust_cov, clust_centroid);
      Eigen::Vector4f plane_params;
      
      EIGEN_ALIGN16 Eigen::Vector3f::Scalar eigen_value;
      EIGEN_ALIGN16 Eigen::Vector3f eigen_vector;
      pcl::eigen33 (clust_cov, eigen_value, eigen_vector);
      plane_params[0] = eigen_vector[0];
      plane_params[1] = eigen_vector[1];
      plane_params[2] = eigen_vector[2];
      plane_params[3] = 0;
      plane_params[3] = -1 * plane_params.dot (clust_centroid);

      vp -= clust_centroid;
      float cos_theta = vp.dot (plane_params);
      if (cos_theta < 0)
      {
        plane_params *= -1;
        plane_params[3] = 0;
        plane_params[3] = -1 * plane_params.dot (clust_centroid);
      }
      
      // Compute the curvature surface change
      float curvature;
      float eig_sum = clust_cov.coeff (0) + clust_cov.coeff (4) + clust_cov.coeff (8);
      if (eig_sum != 0)
        curvature = fabsf (eigen_value / eig_sum);
      else
        curvature = 0;

      if (curvature < maximum_curvature_)
      {
        model.values[0] = plane_params[0];
        model.values[1] = plane_params[1];
        model.values[2] = plane_params[2];
        model.values[3] = plane_params[3];
        model_coefficients.push_back (model);
        inlier_indices.push_back (label_indices[i]);
        centroids.push_back (clust_centroid);
        covariances.push_back (clust_cov);
      }
    }
  }
  deinitCompute ();
}

//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template<typename PointT, typename PointNT, typename PointLT> void
pcl::OrganizedMultiPlaneSegmentation<PointT, PointNT, PointLT>::segment (std::vector<PlanarRegion<PointT>, Eigen::aligned_allocator<PlanarRegion<PointT> > >& regions)
{
  std::vector<ModelCoefficients> model_coefficients;
  std::vector<PointIndices> inlier_indices;  
  PointCloudLPtr labels (new PointCloudL);
  std::vector<pcl::PointIndices> label_indices;
  std::vector<pcl::PointIndices> boundary_indices;
  pcl::PointCloud<PointT> boundary_cloud;
  std::vector<Eigen::Vector4f, Eigen::aligned_allocator<Eigen::Vector4f> > centroids;
  std::vector <Eigen::Matrix3f, Eigen::aligned_allocator<Eigen::Matrix3f> > covariances;
  segment (model_coefficients, inlier_indices, centroids, covariances, *labels, label_indices);
  regions.resize (model_coefficients.size ());
  boundary_indices.resize (model_coefficients.size ());
  
  for (size_t i = 0; i < model_coefficients.size (); i++)
  {
    boundary_cloud.resize (0);
    pcl::OrganizedConnectedComponentSegmentation<PointT,PointLT>::findLabeledRegionBoundary (inlier_indices[i].indices[0], labels, boundary_indices[i]);
    boundary_cloud.points.resize (boundary_indices[i].indices.size ());
    for (unsigned j = 0; j < boundary_indices[i].indices.size (); j++)
      boundary_cloud.points[j] = input_->points[boundary_indices[i].indices[j]];
    
    Eigen::Vector3f centroid = Eigen::Vector3f (centroids[i][0],centroids[i][1],centroids[i][2]);
    Eigen::Vector4f model = Eigen::Vector4f (model_coefficients[i].values[0],
                                             model_coefficients[i].values[1],
                                             model_coefficients[i].values[2],
                                             model_coefficients[i].values[3]);
    regions[i] = PlanarRegion<PointT> (centroid,
                                       covariances[i], 
                                       static_cast<unsigned int> (inlier_indices[i].indices.size ()),
                                       boundary_cloud.points,
                                       model);
  }
}

//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template<typename PointT, typename PointNT, typename PointLT> void
pcl::OrganizedMultiPlaneSegmentation<PointT, PointNT, PointLT>::segmentAndRefine (std::vector<PlanarRegion<PointT>, Eigen::aligned_allocator<PlanarRegion<PointT> > >& regions)
{
  std::vector<ModelCoefficients> model_coefficients;
  std::vector<PointIndices> inlier_indices;  
  PointCloudLPtr labels (new PointCloudL);
  std::vector<pcl::PointIndices> label_indices;
  std::vector<pcl::PointIndices> boundary_indices;
  pcl::PointCloud<PointT> boundary_cloud;
  std::vector<Eigen::Vector4f, Eigen::aligned_allocator<Eigen::Vector4f> > centroids;
  std::vector <Eigen::Matrix3f, Eigen::aligned_allocator<Eigen::Matrix3f> > covariances;
  segment (model_coefficients, inlier_indices, centroids, covariances, *labels, label_indices);
  refine (model_coefficients, inlier_indices, centroids, covariances, labels, label_indices);
  regions.resize (model_coefficients.size ());
  boundary_indices.resize (model_coefficients.size ());

  for (size_t i = 0; i < model_coefficients.size (); i++)
  {
    boundary_cloud.resize (0);
    int max_inlier_idx = static_cast<int> (inlier_indices[i].indices.size ()) - 1;
    pcl::OrganizedConnectedComponentSegmentation<PointT,PointLT>::findLabeledRegionBoundary (inlier_indices[i].indices[max_inlier_idx], labels, boundary_indices[i]);
    boundary_cloud.points.resize (boundary_indices[i].indices.size ());
    for (unsigned j = 0; j < boundary_indices[i].indices.size (); j++)
      boundary_cloud.points[j] = input_->points[boundary_indices[i].indices[j]];
    
    Eigen::Vector3f centroid = Eigen::Vector3f (centroids[i][0],centroids[i][1],centroids[i][2]);
    Eigen::Vector4f model = Eigen::Vector4f (model_coefficients[i].values[0],
                                             model_coefficients[i].values[1],
                                             model_coefficients[i].values[2],
                                             model_coefficients[i].values[3]);

    Eigen::Vector3f vp (0.0, 0.0, 0.0);
    if (project_points_)
      boundary_cloud = projectToPlaneFromViewpoint (boundary_cloud, model, centroid, vp);

    regions[i] = PlanarRegion<PointT> (centroid,
                                       covariances[i], 
                                       static_cast<unsigned int> (inlier_indices[i].indices.size ()),
                                       boundary_cloud.points,
                                       model);
  }
}

//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template<typename PointT, typename PointNT, typename PointLT> void
pcl::OrganizedMultiPlaneSegmentation<PointT, PointNT, PointLT>::segmentAndRefine (std::vector<PlanarRegion<PointT>, Eigen::aligned_allocator<PlanarRegion<PointT> > >& regions,
                                                                                  std::vector<ModelCoefficients>& model_coefficients,
                                                                                  std::vector<PointIndices>& inlier_indices,
                                                                                  PointCloudLPtr& labels,
                                                                                  std::vector<pcl::PointIndices>& label_indices,
                                                                                  std::vector<pcl::PointIndices>& boundary_indices)
{
  pcl::PointCloud<PointT> boundary_cloud;
  std::vector<Eigen::Vector4f, Eigen::aligned_allocator<Eigen::Vector4f> > centroids;
  std::vector <Eigen::Matrix3f, Eigen::aligned_allocator<Eigen::Matrix3f> > covariances;
  segment (model_coefficients, inlier_indices, centroids, covariances, *labels, label_indices);
  refine (model_coefficients, inlier_indices, centroids, covariances, labels, label_indices);
  regions.resize (model_coefficients.size ());
  boundary_indices.resize (model_coefficients.size ());
  
  for (size_t i = 0; i < model_coefficients.size (); i++)
  {
    boundary_cloud.resize (0);
    int max_inlier_idx = static_cast<int> (inlier_indices[i].indices.size ()) - 1;
    pcl::OrganizedConnectedComponentSegmentation<PointT,PointLT>::findLabeledRegionBoundary (inlier_indices[i].indices[max_inlier_idx], labels, boundary_indices[i]);
    boundary_cloud.points.resize (boundary_indices[i].indices.size ());
    for (unsigned j = 0; j < boundary_indices[i].indices.size (); j++)
      boundary_cloud.points[j] = input_->points[boundary_indices[i].indices[j]];

    Eigen::Vector3f centroid = Eigen::Vector3f (centroids[i][0],centroids[i][1],centroids[i][2]);
    Eigen::Vector4f model = Eigen::Vector4f (model_coefficients[i].values[0],
                                             model_coefficients[i].values[1],
                                             model_coefficients[i].values[2],
                                             model_coefficients[i].values[3]);

    Eigen::Vector3f vp (0.0, 0.0, 0.0);
    if (project_points_ && boundary_cloud.points.size () > 0)
      boundary_cloud = projectToPlaneFromViewpoint (boundary_cloud, model, centroid, vp);

    regions[i] = PlanarRegion<PointT> (centroid,
                                       covariances[i], 
                                       static_cast<unsigned int> (inlier_indices[i].indices.size ()),
                                       boundary_cloud.points,
                                       model);
  }
}

//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
template<typename PointT, typename PointNT, typename PointLT> void
pcl::OrganizedMultiPlaneSegmentation<PointT, PointNT, PointLT>::refine (std::vector<ModelCoefficients>& model_coefficients, 
                                                                        std::vector<PointIndices>& inlier_indices,
                                                                        std::vector<Eigen::Vector4f, Eigen::aligned_allocator<Eigen::Vector4f> >&,
                                                                        std::vector <Eigen::Matrix3f, Eigen::aligned_allocator<Eigen::Matrix3f> >&,
                                                                        PointCloudLPtr& labels,
                                                                        std::vector<pcl::PointIndices>& label_indices)
{
  //List of labels to grow, and index of model corresponding to each label
  std::vector<bool> grow_labels;
  std::vector<int> label_to_model;
  grow_labels.resize (label_indices.size (), false);
  label_to_model.resize (label_indices.size (), 0);

  for (size_t i = 0; i < model_coefficients.size (); i++)
  {
    int model_label = (*labels)[inlier_indices[i].indices[0]].label;
    label_to_model[model_label] = static_cast<int> (i);
    grow_labels[model_label] = true;
  }
  
  //refinement_compare_->setDistanceThreshold (0.015f, true);
  refinement_compare_->setInputCloud (input_);
  refinement_compare_->setLabels (labels);
  refinement_compare_->setModelCoefficients (model_coefficients);
  refinement_compare_->setRefineLabels (grow_labels);
  refinement_compare_->setLabelToModel (label_to_model);

  //Do a first pass over the image, top to bottom, left to right
  unsigned int current_row = 0;
  unsigned int next_row = labels->width;
  for (size_t rowIdx = 0; rowIdx < labels->height - 1; ++rowIdx, current_row = next_row, next_row += labels->width)
  {

    for (unsigned colIdx = 0; colIdx < labels->width - 1; ++colIdx)
    {
      int current_label = (*labels)[current_row+colIdx].label;
      int right_label = (*labels)[current_row+colIdx+1].label;
      if (current_label < 0 || right_label < 0)
        continue;
      
      //Check right
      //bool test1 = false;
      if (refinement_compare_->compare (current_row+colIdx, current_row+colIdx+1))
      {
        //test1 = true;
        labels->points[current_row+colIdx+1].label = current_label;
        label_indices[current_label].indices.push_back (current_row+colIdx+1);
        inlier_indices[label_to_model[current_label]].indices.push_back (current_row+colIdx+1);
      }
      
      int lower_label = (*labels)[next_row+colIdx].label;
      if (lower_label < 0)
        continue;
      
      //Check down
      if (refinement_compare_->compare (current_row+colIdx, next_row+colIdx))
      {
        labels->points[next_row+colIdx].label = current_label;
        label_indices[current_label].indices.push_back (next_row+colIdx);
        inlier_indices[label_to_model[current_label]].indices.push_back (next_row+colIdx);
      }

    }//col
  }//row

  //Do a second pass over the image
  current_row = labels->width * (labels->height - 1);
  unsigned int prev_row = current_row - labels->width;
  for (size_t rowIdx = 0; rowIdx < labels->height - 1; ++rowIdx, current_row = prev_row, prev_row -= labels->width)
  {
    for (int colIdx = labels->width - 1; colIdx >= 0; --colIdx)
    {
      int current_label = (*labels)[current_row+colIdx].label;
      int left_label    = (*labels)[current_row+colIdx-1].label;
      if (current_label < 0 || left_label < 0)
        continue;

      //Check left
      if (refinement_compare_->compare (current_row+colIdx, current_row+colIdx-1))
      {
        labels->points[current_row+colIdx-1].label = current_label;
        label_indices[current_label].indices.push_back (current_row+colIdx-1);
        inlier_indices[label_to_model[current_label]].indices.push_back (current_row+colIdx-1);
      }
      
      int upper_label    = (*labels)[prev_row+colIdx].label;
      if (upper_label < 0)
        continue;
      //Check up
      if (refinement_compare_->compare (current_row+colIdx, prev_row+colIdx))
      {
        labels->points[prev_row+colIdx].label = current_label;
        label_indices[current_label].indices.push_back (prev_row+colIdx);
        inlier_indices[label_to_model[current_label]].indices.push_back (prev_row+colIdx);
      }
    }//col
  }//row
}

#define PCL_INSTANTIATE_OrganizedMultiPlaneSegmentation(T,NT,LT) template class PCL_EXPORTS pcl::OrganizedMultiPlaneSegmentation<T,NT,LT>;

#endif  // PCL_SEGMENTATION_IMPL_MULTI_PLANE_SEGMENTATION_H_