Point Cloud Library (PCL) 1.15.0
Loading...
Searching...
No Matches
cpc_segmentation.hpp
1/*
2 * Software License Agreement (BSD License)
3 *
4 * Point Cloud Library (PCL) - www.pointclouds.org
5 * Copyright (c) 2014-, Open Perception, 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 the copyright holder(s) 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
38#ifndef PCL_SEGMENTATION_IMPL_CPC_SEGMENTATION_HPP_
39#define PCL_SEGMENTATION_IMPL_CPC_SEGMENTATION_HPP_
40
41#include <pcl/sample_consensus/sac_model_plane.h> // for SampleConsensusModelPlane
42#include <pcl/segmentation/cpc_segmentation.h>
43
44template <typename PointT>
46
47template <typename PointT>
49
50template <typename PointT> void
52{
53 if (supervoxels_set_)
54 {
55 // Calculate for every Edge if the connection is convex or invalid
56 // This effectively performs the segmentation.
57 calculateConvexConnections (sv_adjacency_list_);
58
59 // Correct edge relations using extended convexity definition if k>0
60 applyKconvexity (k_factor_);
61
62 // Determine whether to use cutting planes
63 doGrouping ();
64
65 grouping_data_valid_ = true;
66
67 applyCuttingPlane (max_cuts_);
68
69 // merge small segments
70 mergeSmallSegments ();
71 }
72 else
73 PCL_WARN ("[pcl::CPCSegmentation::segment] WARNING: Call function setInputSupervoxels first. Nothing has been done. \n");
74}
75
76template <typename PointT> void
77pcl::CPCSegmentation<PointT>::applyCuttingPlane (std::uint32_t depth_levels_left)
78{
79 using SegLabel2ClusterMap = std::map<std::uint32_t, pcl::PointCloud<WeightSACPointType>::Ptr>;
80
81 pcl::console::print_info ("Cutting at level %d (maximum %d)\n", max_cuts_ - depth_levels_left + 1, max_cuts_);
82 // stop if we reached the 0 level
83 if (depth_levels_left <= 0)
84 return;
85
86 pcl::IndicesPtr support_indices (new pcl::Indices);
87 SegLabel2ClusterMap seg_to_edge_points_map;
88 std::map<std::uint32_t, std::vector<EdgeID> > seg_to_edgeIDs_map;
89 EdgeIterator edge_itr, edge_itr_end, next_edge;
90 boost::tie (edge_itr, edge_itr_end) = boost::edges (sv_adjacency_list_);
91 for (next_edge = edge_itr; edge_itr != edge_itr_end; edge_itr = next_edge)
92 {
93 next_edge++; // next_edge iterator is necessary, because removing an edge invalidates the iterator to the current edge
94 std::uint32_t source_sv_label = sv_adjacency_list_[boost::source (*edge_itr, sv_adjacency_list_)];
95 std::uint32_t target_sv_label = sv_adjacency_list_[boost::target (*edge_itr, sv_adjacency_list_)];
96
97 std::uint32_t source_segment_label = sv_label_to_seg_label_map_[source_sv_label];
98 std::uint32_t target_segment_label = sv_label_to_seg_label_map_[target_sv_label];
99
100 // do not process edges which already split two segments
101 if (source_segment_label != target_segment_label)
102 continue;
103
104 // if edge has been used for cutting already do not use it again
105 if (sv_adjacency_list_[*edge_itr].used_for_cutting)
106 continue;
107 // get centroids of vertices
108 const pcl::PointXYZRGBA source_centroid = sv_label_to_supervoxel_map_[source_sv_label]->centroid_;
109 const pcl::PointXYZRGBA target_centroid = sv_label_to_supervoxel_map_[target_sv_label]->centroid_;
110
111 // stores the information about the edge cloud (used for the weighted ransac)
112 // we use the normal to express the direction of the connection
113 // we use the intensity to express the normal differences between supervoxel patches. <=0: Convex, >0: Concave
114 WeightSACPointType edge_centroid;
115 edge_centroid.getVector3fMap () = (source_centroid.getVector3fMap () + target_centroid.getVector3fMap ()) / 2;
116
117 // we use the normal to express the direction of the connection!
118 edge_centroid.getNormalVector3fMap () = (target_centroid.getVector3fMap () - source_centroid.getVector3fMap ()).normalized ();
119
120 // we use the intensity to express the normal differences between supervoxel patches. <=0: Convex, >0: Concave
121 edge_centroid.intensity = sv_adjacency_list_[*edge_itr].is_convex ? -sv_adjacency_list_[*edge_itr].normal_difference : sv_adjacency_list_[*edge_itr].normal_difference;
122 if (seg_to_edge_points_map.find (source_segment_label) == seg_to_edge_points_map.end ())
123 {
124 seg_to_edge_points_map[source_segment_label] = pcl::PointCloud<WeightSACPointType>::Ptr (new pcl::PointCloud<WeightSACPointType> ());
125 }
126 seg_to_edge_points_map[source_segment_label]->push_back (edge_centroid);
127 seg_to_edgeIDs_map[source_segment_label].push_back (*edge_itr);
128 }
129 bool cut_found = false;
130 // do the following processing for each segment separately
131 for (const auto &seg_to_edge_points : seg_to_edge_points_map)
132 {
133 // if too small do not process
134 if (seg_to_edge_points.second->size () < min_segment_size_for_cutting_)
135 {
136 continue;
137 }
138
139 std::vector<double> weights;
140 weights.resize (seg_to_edge_points.second->size ());
141 for (std::size_t cp = 0; cp < seg_to_edge_points.second->size (); ++cp)
142 {
143 float& cur_weight = (*seg_to_edge_points.second)[cp].intensity;
144 cur_weight = cur_weight < concavity_tolerance_threshold_ ? 0 : 1;
145 weights[cp] = cur_weight;
146 }
147
148 pcl::PointCloud<WeightSACPointType>::Ptr edge_cloud_cluster = seg_to_edge_points.second;
150
151 WeightedRandomSampleConsensus weight_sac (model_p, seed_resolution_, true);
152
153 weight_sac.setWeights (weights, use_directed_weights_);
154 weight_sac.setMaxIterations (ransac_itrs_);
155
156 // if not enough inliers are found
157 if (!weight_sac.computeModel ())
158 {
159 continue;
160 }
161
162 Eigen::VectorXf model_coefficients;
163 weight_sac.getModelCoefficients (model_coefficients);
164
165 model_coefficients[3] += std::numeric_limits<float>::epsilon ();
166
167 weight_sac.getInliers (*support_indices);
168
169 // the support_indices which are actually cut (if not locally constrain: cut_support_indices = support_indices
170 pcl::Indices cut_support_indices;
171
172 if (use_local_constrains_)
173 {
174 Eigen::Vector3f plane_normal (model_coefficients[0], model_coefficients[1], model_coefficients[2]);
175 // Cut the connections.
176 // We only iterate through the points which are within the support (when we are local, otherwise all points in the segment).
177 // We also just actually cut when the edge goes through the plane. This is why we check the planedistance
178 std::vector<pcl::PointIndices> cluster_indices;
181 tree->setInputCloud (edge_cloud_cluster);
182 euclidean_clusterer.setClusterTolerance (seed_resolution_);
183 euclidean_clusterer.setMinClusterSize (1);
184 euclidean_clusterer.setMaxClusterSize (25000);
185 euclidean_clusterer.setSearchMethod (tree);
186 euclidean_clusterer.setInputCloud (edge_cloud_cluster);
187 euclidean_clusterer.setIndices (support_indices);
188 euclidean_clusterer.extract (cluster_indices);
189// sv_adjacency_list_[seg_to_edgeID_map[seg_to_edge_points.first][point_index]].used_for_cutting = true;
190
191 for (const auto &cluster_index : cluster_indices)
192 {
193 // get centroids of vertices
194 float cluster_score = 0;
195// std::cout << "Cluster has " << cluster_indices[cc].indices.size () << " points" << std::endl;
196 for (const auto &current_index : cluster_index.indices)
197 {
198 double index_score = weights[current_index];
199 if (use_directed_weights_)
200 index_score *= 1.414 * (std::abs (plane_normal.dot (edge_cloud_cluster->at (current_index).getNormalVector3fMap ())));
201 cluster_score += index_score;
202 }
203 // check if the score is below the threshold. If that is the case this segment should not be split
204 cluster_score /= cluster_index.indices.size ();
205// std::cout << "Cluster score: " << cluster_score << std::endl;
206 if (cluster_score >= min_cut_score_)
207 {
208 cut_support_indices.insert (cut_support_indices.end (), cluster_index.indices.begin (), cluster_index.indices.end ());
209 }
210 }
211 if (cut_support_indices.empty ())
212 {
213// std::cout << "Could not find planes which exceed required minimum score (threshold " << min_cut_score_ << "), not cutting" << std::endl;
214 continue;
215 }
216 }
217 else
218 {
219 double current_score = weight_sac.getBestScore ();
220 cut_support_indices = *support_indices;
221 // check if the score is below the threshold. If that is the case this segment should not be split
222 if (current_score < min_cut_score_)
223 {
224// std::cout << "Score too low, no cutting" << std::endl;
225 continue;
226 }
227 }
228
229 int number_connections_cut = 0;
230 for (const auto &point_index : cut_support_indices)
231 {
232 if (use_clean_cutting_)
233 {
234 // skip edges where both centroids are on one side of the cutting plane
235 std::uint32_t source_sv_label = sv_adjacency_list_[boost::source (seg_to_edgeIDs_map[seg_to_edge_points.first][point_index], sv_adjacency_list_)];
236 std::uint32_t target_sv_label = sv_adjacency_list_[boost::target (seg_to_edgeIDs_map[seg_to_edge_points.first][point_index], sv_adjacency_list_)];
237 // get centroids of vertices
238 const pcl::PointXYZRGBA source_centroid = sv_label_to_supervoxel_map_[source_sv_label]->centroid_;
239 const pcl::PointXYZRGBA target_centroid = sv_label_to_supervoxel_map_[target_sv_label]->centroid_;
240 // this makes a clean cut
241 if (pcl::pointToPlaneDistanceSigned (source_centroid, model_coefficients) * pcl::pointToPlaneDistanceSigned (target_centroid, model_coefficients) > 0)
242 {
243 continue;
244 }
245 }
246 sv_adjacency_list_[seg_to_edgeIDs_map[seg_to_edge_points.first][point_index]].used_for_cutting = true;
247 if (sv_adjacency_list_[seg_to_edgeIDs_map[seg_to_edge_points.first][point_index]].is_valid)
248 {
249 ++number_connections_cut;
250 sv_adjacency_list_[seg_to_edgeIDs_map[seg_to_edge_points.first][point_index]].is_valid = false;
251 }
252 }
253// std::cout << "We cut " << number_connections_cut << " connections" << std::endl;
254 if (number_connections_cut > 0)
255 cut_found = true;
256 }
257
258 // if not cut has been performed we can stop the recursion
259 if (cut_found)
260 {
261 doGrouping ();
262 --depth_levels_left;
263 applyCuttingPlane (depth_levels_left);
264 }
265 else
266 pcl::console::print_info ("Could not find any more cuts, stopping recursion\n");
267}
268
269/******************************************* Directional weighted RANSAC definitions ******************************************************************/
270
271
272template <typename PointT> bool
274{
275 // Warn and exit if no threshold was set
276 if (threshold_ == std::numeric_limits<double>::max ())
277 {
278 PCL_ERROR ("[pcl::CPCSegmentation<PointT>::WeightedRandomSampleConsensus::computeModel] No threshold set!\n");
279 return (false);
280 }
281
282 iterations_ = 0;
283 best_score_ = -std::numeric_limits<double>::max ();
284
285 pcl::Indices selection;
286 Eigen::VectorXf model_coefficients;
287
288 unsigned skipped_count = 0;
289 // suppress infinite loops by just allowing 10 x maximum allowed iterations for invalid model parameters!
290 const unsigned max_skip = max_iterations_ * 10;
291
292 // Iterate
293 while (iterations_ < max_iterations_ && skipped_count < max_skip)
294 {
295 // Get X samples which satisfy the model criteria and which have a weight > 0
296 sac_model_->setIndices (model_pt_indices_);
297 sac_model_->getSamples (iterations_, selection);
298
299 if (selection.empty ())
300 {
301 PCL_ERROR ("[pcl::CPCSegmentation<PointT>::WeightedRandomSampleConsensus::computeModel] No samples could be selected!\n");
302 break;
303 }
304
305 // Search for inliers in the point cloud for the current plane model M
306 if (!sac_model_->computeModelCoefficients (selection, model_coefficients))
307 {
308 //++iterations_;
309 ++skipped_count;
310 continue;
311 }
312 // weight distances to get the score (only using connected inliers)
313 sac_model_->setIndices (full_cloud_pt_indices_);
314
315 pcl::IndicesPtr current_inliers (new pcl::Indices);
316 sac_model_->selectWithinDistance (model_coefficients, threshold_, *current_inliers);
317 double current_score = 0;
318 Eigen::Vector3f plane_normal (model_coefficients[0], model_coefficients[1], model_coefficients[2]);
319 for (const auto &current_index : *current_inliers)
320 {
321 double index_score = weights_[current_index];
322 if (use_directed_weights_)
323 // the sqrt(2) factor was used in the paper and was meant for making the scores better comparable between directed and undirected weights
324 index_score *= 1.414 * (std::abs (plane_normal.dot (point_cloud_ptr_->at (current_index).getNormalVector3fMap ())));
325
326 current_score += index_score;
327 }
328 // normalize by the total number of inliers
329 current_score /= current_inliers->size ();
330
331 // Better match ?
332 if (current_score > best_score_)
333 {
334 best_score_ = current_score;
335 // Save the current model/inlier/coefficients selection as being the best so far
336 model_ = selection;
337 model_coefficients_ = model_coefficients;
338 }
339
340 ++iterations_;
341 PCL_DEBUG ("[pcl::CPCSegmentation<PointT>::WeightedRandomSampleConsensus::computeModel] Trial %d (max %d): score is %f (best is: %f so far).\n", iterations_, max_iterations_, current_score, best_score_);
342 if (iterations_ > max_iterations_)
343 {
344 PCL_DEBUG ("[pcl::CPCSegmentation<PointT>::WeightedRandomSampleConsensus::computeModel] RANSAC reached the maximum number of trials.\n");
345 break;
346 }
347 }
348// std::cout << "Took us " << iterations_ - 1 << " iterations" << std::endl;
349 PCL_DEBUG ("[pcl::CPCSegmentation<PointT>::WeightedRandomSampleConsensus::computeModel] Model: %lu size, %f score.\n", model_.size (), best_score_);
350
351 if (model_.empty ())
352 {
353 inliers_.clear ();
354 return (false);
355 }
356
357 // Get the set of inliers that correspond to the best model found so far
358 sac_model_->selectWithinDistance (model_coefficients_, threshold_, inliers_);
359 return (true);
360}
361
362#endif // PCL_SEGMENTATION_IMPL_CPC_SEGMENTATION_HPP_
A segmentation algorithm partitioning a supervoxel graph.
void segment()
Merge supervoxels using cuts through local convexities.
~CPCSegmentation() override
EuclideanClusterExtraction represents a segmentation class for cluster extraction in an Euclidean sen...
void extract(std::vector< PointIndices > &clusters)
Cluster extraction in a PointCloud given by <setInputCloud (), setIndices ()>
void setClusterTolerance(double tolerance)
Set the spatial cluster tolerance as a measure in the L2 Euclidean space.
void setSearchMethod(const KdTreePtr &tree)
Provide a pointer to the search object.
void setMaxClusterSize(pcl::uindex_t max_cluster_size)
Set the maximum number of points that a cluster needs to contain in order to be considered valid.
void setMinClusterSize(pcl::uindex_t min_cluster_size)
Set the minimum number of points that a cluster needs to contain in order to be considered valid.
virtual void setInputCloud(const PointCloudConstPtr &cloud)
Provide a pointer to the input dataset.
Definition pcl_base.hpp:65
virtual void setIndices(const IndicesPtr &indices)
Provide a pointer to the vector of indices that represents the input data.
Definition pcl_base.hpp:72
PointCloud represents the base class in PCL for storing collections of 3D points.
const PointT & at(int column, int row) const
Obtain the point given by the (column, row) coordinates.
shared_ptr< PointCloud< PointT > > Ptr
SampleConsensusModelPlane defines a model for 3D plane segmentation.
shared_ptr< SampleConsensusModelPlane< PointT > > Ptr
search::KdTree is a wrapper class which inherits the pcl::KdTree class for performing search function...
Definition kdtree.h:62
shared_ptr< KdTree< PointT, Tree > > Ptr
Definition kdtree.h:75
double pointToPlaneDistanceSigned(const Point &p, double a, double b, double c, double d)
Get the distance from a point to a plane (signed) defined by ax+by+cz+d=0.
PCL_EXPORTS void print_info(const char *format,...)
Print an info message on stream with colors.
int cp(int from, int to)
Returns field copy operation code.
Definition repacks.hpp:54
IndicesAllocator<> Indices
Type used for indices in PCL.
Definition types.h:133
shared_ptr< Indices > IndicesPtr
Definition pcl_base.h:58
A point structure representing Euclidean xyz coordinates, and the RGBA color.