Point Cloud Library (PCL)  1.14.1-dev
hv_papazov.hpp
1 /*
2  * Software License Agreement (BSD License)
3  *
4  * Point Cloud Library (PCL) - www.pointclouds.org
5  * Copyright (c) 2010-2011, Willow Garage, Inc.
6  *
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  * * Redistributions of source code must retain the above copyright
14  * notice, this list of conditions and the following disclaimer.
15  * * Redistributions in binary form must reproduce the above
16  * copyright notice, this list of conditions and the following
17  * disclaimer in the documentation and/or other materials provided
18  * with the distribution.
19  * * Neither the name of Willow Garage, Inc. nor the names of its
20  * contributors may be used to endorse or promote products derived
21  * from this software without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
26  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
27  * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
28  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
29  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
30  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
31  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
33  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #pragma once
38 #include <pcl/recognition/hv/hv_papazov.h>
39 
40 ///////////////////////////////////////////////////////////////////////////////////////////////////
41 template<typename ModelT, typename SceneT>
42  void
44  {
45 
46  //clear stuff
47  recognition_models_.clear ();
48  graph_id_model_map_.clear ();
49  conflict_graph_.clear ();
50  explained_by_RM_.clear ();
51  points_explained_by_rm_.clear ();
52 
53  // initialize mask...
54  mask_.resize (complete_models_.size ());
55  for (std::size_t i = 0; i < complete_models_.size (); i++)
56  mask_[i] = true;
57 
58  // initialize explained_by_RM
59  explained_by_RM_.resize (scene_cloud_downsampled_->size ());
60  points_explained_by_rm_.resize (scene_cloud_downsampled_->size ());
61 
62  // initialize model
63  for (std::size_t m = 0; m < complete_models_.size (); m++)
64  {
65  RecognitionModelPtr recog_model (new RecognitionModel);
66  // voxelize model cloud
67  recog_model->cloud_.reset (new pcl::PointCloud<ModelT>);
68  recog_model->complete_cloud_.reset (new pcl::PointCloud<ModelT>);
69  recog_model->id_ = static_cast<int> (m);
70 
71  pcl::VoxelGrid<ModelT> voxel_grid;
72  voxel_grid.setInputCloud (visible_models_[m]);
73  voxel_grid.setLeafSize (resolution_, resolution_, resolution_);
74  voxel_grid.filter (*(recog_model->cloud_));
75 
76  pcl::VoxelGrid<ModelT> voxel_grid_complete;
77  voxel_grid_complete.setInputCloud (complete_models_[m]);
78  voxel_grid_complete.setLeafSize (resolution_, resolution_, resolution_);
79  voxel_grid_complete.filter (*(recog_model->complete_cloud_));
80 
81  std::vector<int> explained_indices;
82  std::vector<int> outliers;
83  pcl::Indices nn_indices;
84  std::vector<float> nn_distances;
85 
86  for (std::size_t i = 0; i < recog_model->cloud_->size (); i++)
87  {
88  if (!scene_downsampled_tree_->radiusSearch ((*recog_model->cloud_)[i], inliers_threshold_, nn_indices, nn_distances,
89  std::numeric_limits<int>::max ()))
90  {
91  outliers.push_back (static_cast<int> (i));
92  }
93  else
94  {
95  for (std::size_t k = 0; k < nn_distances.size (); k++)
96  {
97  explained_indices.push_back (nn_indices[k]); //nn_indices[k] points to the scene
98  }
99  }
100  }
101 
102  std::sort (explained_indices.begin (), explained_indices.end ());
103  explained_indices.erase (std::unique (explained_indices.begin (), explained_indices.end ()), explained_indices.end ());
104 
105  recog_model->bad_information_ = static_cast<int> (outliers.size ());
106 
107  if ((static_cast<float> (recog_model->bad_information_) / static_cast<float> (recog_model->complete_cloud_->size ()))
108  <= penalty_threshold_ && (static_cast<float> (explained_indices.size ())
109  / static_cast<float> (recog_model->complete_cloud_->size ())) >= support_threshold_)
110  {
111  recog_model->explained_ = explained_indices;
112  recognition_models_.push_back (recog_model);
113 
114  // update explained_by_RM_, add 1
115  for (const int &explained_index : explained_indices)
116  {
117  explained_by_RM_[explained_index]++;
118  points_explained_by_rm_[explained_index].push_back (recog_model);
119  }
120  }
121  else
122  {
123  mask_[m] = false; // the model didn't survive the sequential check...
124  }
125  }
126  }
127 
128 ///////////////////////////////////////////////////////////////////////////////////////////////////
129 template<typename ModelT, typename SceneT>
130  void
132  {
133  // iterate over all vertices of the graph and check if they have a better neighbour, then remove that vertex
134  using VertexIterator = typename boost::graph_traits<Graph>::vertex_iterator;
135  VertexIterator vi, vi_end;
136  boost::tie (vi, vi_end) = boost::vertices (conflict_graph_);
137 
138  for (auto next = vi; next != vi_end; next++)
139  {
140  const typename Graph::vertex_descriptor v = boost::vertex (*next, conflict_graph_);
141  typename boost::graph_traits<Graph>::adjacency_iterator ai;
142  typename boost::graph_traits<Graph>::adjacency_iterator ai_end;
143 
144  auto current = std::static_pointer_cast<RecognitionModel> (graph_id_model_map_[static_cast<int>(v)]);
145 
146  bool a_better_one = false;
147  for (boost::tie (ai, ai_end) = boost::adjacent_vertices (v, conflict_graph_); (ai != ai_end) && !a_better_one; ++ai)
148  {
149  auto neighbour = std::static_pointer_cast<RecognitionModel> (graph_id_model_map_[static_cast<int>(*ai)]);
150  if ((neighbour->explained_.size () >= current->explained_.size ()) && mask_[neighbour->id_])
151  {
152  a_better_one = true;
153  }
154  }
155 
156  if (a_better_one)
157  {
158  mask_[current->id_] = false;
159  }
160  }
161  }
162 
163 ///////////////////////////////////////////////////////////////////////////////////////////////////
164 template<typename ModelT, typename SceneT>
165  void
167  {
168  // create vertices for the graph
169  for (std::size_t i = 0; i < (recognition_models_.size ()); i++)
170  {
171  const typename Graph::vertex_descriptor v = boost::add_vertex (recognition_models_[i], conflict_graph_);
172  graph_id_model_map_[static_cast<int>(v)] = std::static_pointer_cast<RecognitionModel> (recognition_models_[i]);
173  }
174 
175  // iterate over the remaining models and check for each one if there is a conflict with another one
176  for (std::size_t i = 0; i < recognition_models_.size (); i++)
177  {
178  for (std::size_t j = i; j < recognition_models_.size (); j++)
179  {
180  if (i != j)
181  {
182  float n_conflicts = 0.f;
183  // count scene points explained by both models
184  for (std::size_t k = 0; k < explained_by_RM_.size (); k++)
185  {
186  if (explained_by_RM_[k] > 1)
187  {
188  // this point could be a conflict
189  bool i_found = false;
190  bool j_found = false;
191  bool both_found = false;
192  for (std::size_t kk = 0; (kk < points_explained_by_rm_[k].size ()) && !both_found; kk++)
193  {
194  if (points_explained_by_rm_[k][kk]->id_ == recognition_models_[i]->id_)
195  i_found = true;
196 
197  if (points_explained_by_rm_[k][kk]->id_ == recognition_models_[j]->id_)
198  j_found = true;
199 
200  if (i_found && j_found)
201  both_found = true;
202  }
203 
204  if (both_found)
205  n_conflicts += 1.f;
206  }
207  }
208 
209  // check if number of points is big enough to create a conflict
210  bool add_conflict = false;
211  add_conflict = ((n_conflicts / static_cast<float> (recognition_models_[i]->complete_cloud_->size ())) > conflict_threshold_size_)
212  || ((n_conflicts / static_cast<float> (recognition_models_[j]->complete_cloud_->size ())) > conflict_threshold_size_);
213 
214  if (add_conflict)
215  {
216  boost::add_edge (i, j, conflict_graph_);
217  }
218  }
219  }
220  }
221  }
222 
223 ///////////////////////////////////////////////////////////////////////////////////////////////////
224 template<typename ModelT, typename SceneT>
225  void
227  {
228  initialize ();
229  buildConflictGraph ();
230  nonMaximaSuppresion ();
231  }
232 
233 #define PCL_INSTANTIATE_PapazovHV(T1,T2) template class PCL_EXPORTS pcl::PapazovHV<T1,T2>;
void filter(PointCloud &output)
Calls the filtering method and returns the filtered dataset in output.
Definition: filter.h:121
virtual void setInputCloud(const PointCloudConstPtr &cloud)
Provide a pointer to the input dataset.
Definition: pcl_base.hpp:65
A hypothesis verification method proposed in "An Efficient RANSAC for 3D Object Recognition in Noisy ...
Definition: hv_papazov.h:56
void verify() override
Definition: hv_papazov.hpp:226
VoxelGrid assembles a local 3D grid over a given PointCloud, and downsamples + filters the data.
Definition: voxel_grid.h:210
void setLeafSize(const Eigen::Vector4f &leaf_size)
Set the voxel grid leaf size.
Definition: voxel_grid.h:247
IndicesAllocator<> Indices
Type used for indices in PCL.
Definition: types.h:133