Point Cloud Library (PCL) 1.13.0
Loading...
Searching...
No Matches
correspondence_estimation_backprojection.hpp
1/*
2 * Software License Agreement (BSD License)
3 *
4 * Point Cloud Library (PCL) - www.pointclouds.org
5 * Copyright (c) 2012-, 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 * $Id$
37 *
38 */
39
40#ifndef PCL_REGISTRATION_IMPL_CORRESPONDENCE_ESTIMATION_BACK_PROJECTION_HPP_
41#define PCL_REGISTRATION_IMPL_CORRESPONDENCE_ESTIMATION_BACK_PROJECTION_HPP_
42
43#include <pcl/common/copy_point.h>
44
45namespace pcl {
46
47namespace registration {
48
49template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
50bool
53{
54 if (!source_normals_ || !target_normals_) {
55 PCL_WARN("[pcl::registration::%s::initCompute] Datasets containing normals for "
56 "source/target have not been given!\n",
57 getClassName().c_str());
58 return (false);
59 }
60
61 return (
63}
64
65///////////////////////////////////////////////////////////////////////////////////////////
66template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
67void
69 determineCorrespondences(pcl::Correspondences& correspondences, double max_distance)
70{
71 if (!initCompute())
72 return;
73
74 correspondences.resize(indices_->size());
75
76 pcl::Indices nn_indices(k_);
77 std::vector<float> nn_dists(k_);
78
79 int min_index = 0;
80
82 unsigned int nr_valid_correspondences = 0;
83
84 // Check if the template types are the same. If true, avoid a copy.
85 // Both point types MUST be registered using the POINT_CLOUD_REGISTER_POINT_STRUCT
86 // macro!
87 if (isSamePointType<PointSource, PointTarget>()) {
88 PointTarget pt;
89 // Iterate over the input set of source indices
90 for (const auto& idx_i : (*indices_)) {
91 tree_->nearestKSearch((*input_)[idx_i], k_, nn_indices, nn_dists);
92
93 // Among the K nearest neighbours find the one with minimum perpendicular distance
94 // to the normal
95 float min_dist = std::numeric_limits<float>::max();
96
97 // Find the best correspondence
98 for (std::size_t j = 0; j < nn_indices.size(); j++) {
99 float cos_angle = (*source_normals_)[idx_i].normal_x *
100 (*target_normals_)[nn_indices[j]].normal_x +
101 (*source_normals_)[idx_i].normal_y *
102 (*target_normals_)[nn_indices[j]].normal_y +
103 (*source_normals_)[idx_i].normal_z *
104 (*target_normals_)[nn_indices[j]].normal_z;
105 float dist = nn_dists[j] * (2.0f - cos_angle * cos_angle);
106
107 if (dist < min_dist) {
108 min_dist = dist;
109 min_index = static_cast<int>(j);
110 }
111 }
112 if (min_dist > max_distance)
113 continue;
114
115 corr.index_query = idx_i;
116 corr.index_match = nn_indices[min_index];
117 corr.distance = nn_dists[min_index]; // min_dist;
118 correspondences[nr_valid_correspondences++] = corr;
119 }
120 }
121 else {
122 PointTarget pt;
123
124 // Iterate over the input set of source indices
125 for (const auto& idx_i : (*indices_)) {
126 tree_->nearestKSearch((*input_)[idx_i], k_, nn_indices, nn_dists);
127
128 // Among the K nearest neighbours find the one with minimum perpendicular distance
129 // to the normal
130 float min_dist = std::numeric_limits<float>::max();
131
132 // Find the best correspondence
133 for (std::size_t j = 0; j < nn_indices.size(); j++) {
134 PointSource pt_src;
135 // Copy the source data to a target PointTarget format so we can search in the
136 // tree
137 copyPoint((*input_)[idx_i], pt_src);
138
139 float cos_angle = (*source_normals_)[idx_i].normal_x *
140 (*target_normals_)[nn_indices[j]].normal_x +
141 (*source_normals_)[idx_i].normal_y *
142 (*target_normals_)[nn_indices[j]].normal_y +
143 (*source_normals_)[idx_i].normal_z *
144 (*target_normals_)[nn_indices[j]].normal_z;
145 float dist = nn_dists[j] * (2.0f - cos_angle * cos_angle);
146
147 if (dist < min_dist) {
148 min_dist = dist;
149 min_index = static_cast<int>(j);
150 }
151 }
152 if (min_dist > max_distance)
153 continue;
154
155 corr.index_query = idx_i;
156 corr.index_match = nn_indices[min_index];
157 corr.distance = nn_dists[min_index]; // min_dist;
158 correspondences[nr_valid_correspondences++] = corr;
159 }
160 }
161 correspondences.resize(nr_valid_correspondences);
162 deinitCompute();
163}
164
165template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
166void
169 double max_distance)
170{
171 if (!initCompute())
172 return;
173
174 // Set the internal point representation of choice
175 if (!initComputeReciprocal())
176 return;
177
178 correspondences.resize(indices_->size());
179
180 pcl::Indices nn_indices(k_);
181 std::vector<float> nn_dists(k_);
182 pcl::Indices index_reciprocal(1);
183 std::vector<float> distance_reciprocal(1);
184
185 int min_index = 0;
186
188 unsigned int nr_valid_correspondences = 0;
189 int target_idx = 0;
190
191 // Check if the template types are the same. If true, avoid a copy.
192 // Both point types MUST be registered using the POINT_CLOUD_REGISTER_POINT_STRUCT
193 // macro!
194 if (isSamePointType<PointSource, PointTarget>()) {
195 PointTarget pt;
196 // Iterate over the input set of source indices
197 for (const auto& idx_i : (*indices_)) {
198 tree_->nearestKSearch((*input_)[idx_i], k_, nn_indices, nn_dists);
199
200 // Among the K nearest neighbours find the one with minimum perpendicular distance
201 // to the normal
202 float min_dist = std::numeric_limits<float>::max();
203
204 // Find the best correspondence
205 for (std::size_t j = 0; j < nn_indices.size(); j++) {
206 float cos_angle = (*source_normals_)[idx_i].normal_x *
207 (*target_normals_)[nn_indices[j]].normal_x +
208 (*source_normals_)[idx_i].normal_y *
209 (*target_normals_)[nn_indices[j]].normal_y +
210 (*source_normals_)[idx_i].normal_z *
211 (*target_normals_)[nn_indices[j]].normal_z;
212 float dist = nn_dists[j] * (2.0f - cos_angle * cos_angle);
213
214 if (dist < min_dist) {
215 min_dist = dist;
216 min_index = static_cast<int>(j);
217 }
218 }
219 if (min_dist > max_distance)
220 continue;
221
222 // Check if the correspondence is reciprocal
223 target_idx = nn_indices[min_index];
224 tree_reciprocal_->nearestKSearch(
225 (*target_)[target_idx], 1, index_reciprocal, distance_reciprocal);
226
227 if (idx_i != index_reciprocal[0])
228 continue;
229
230 corr.index_query = idx_i;
231 corr.index_match = nn_indices[min_index];
232 corr.distance = nn_dists[min_index]; // min_dist;
233 correspondences[nr_valid_correspondences++] = corr;
234 }
235 }
236 else {
237 PointTarget pt;
238
239 // Iterate over the input set of source indices
240 for (const auto& idx_i : (*indices_)) {
241 tree_->nearestKSearch((*input_)[idx_i], k_, nn_indices, nn_dists);
242
243 // Among the K nearest neighbours find the one with minimum perpendicular distance
244 // to the normal
245 float min_dist = std::numeric_limits<float>::max();
246
247 // Find the best correspondence
248 for (std::size_t j = 0; j < nn_indices.size(); j++) {
249 PointSource pt_src;
250 // Copy the source data to a target PointTarget format so we can search in the
251 // tree
252 copyPoint((*input_)[idx_i], pt_src);
253
254 float cos_angle = (*source_normals_)[idx_i].normal_x *
255 (*target_normals_)[nn_indices[j]].normal_x +
256 (*source_normals_)[idx_i].normal_y *
257 (*target_normals_)[nn_indices[j]].normal_y +
258 (*source_normals_)[idx_i].normal_z *
259 (*target_normals_)[nn_indices[j]].normal_z;
260 float dist = nn_dists[j] * (2.0f - cos_angle * cos_angle);
261
262 if (dist < min_dist) {
263 min_dist = dist;
264 min_index = static_cast<int>(j);
265 }
266 }
267 if (min_dist > max_distance)
268 continue;
269
270 // Check if the correspondence is reciprocal
271 target_idx = nn_indices[min_index];
272 tree_reciprocal_->nearestKSearch(
273 (*target_)[target_idx], 1, index_reciprocal, distance_reciprocal);
274
275 if (idx_i != index_reciprocal[0])
276 continue;
277
278 corr.index_query = idx_i;
279 corr.index_match = nn_indices[min_index];
280 corr.distance = nn_dists[min_index]; // min_dist;
281 correspondences[nr_valid_correspondences++] = corr;
282 }
283 }
284 correspondences.resize(nr_valid_correspondences);
285 deinitCompute();
286}
287
288} // namespace registration
289} // namespace pcl
290
291#endif // PCL_REGISTRATION_IMPL_CORRESPONDENCE_ESTIMATION_BACK_PROJECTION_HPP_
virtual void determineReciprocalCorrespondences(pcl::Correspondences &correspondences, double max_distance=std::numeric_limits< double >::max())
Determine the reciprocal correspondences between input and target cloud.
void determineCorrespondences(pcl::Correspondences &correspondences, double max_distance=std::numeric_limits< double >::max())
Determine the correspondences between input and target cloud.
Abstract CorrespondenceEstimationBase class.
void copyPoint(const PointInT &point_in, PointOutT &point_out)
Copy the fields of a source point into a target point.
std::vector< pcl::Correspondence, Eigen::aligned_allocator< pcl::Correspondence > > Correspondences
IndicesAllocator<> Indices
Type used for indices in PCL.
Definition types.h:133
Correspondence represents a match between two entities (e.g., points, descriptors,...
index_t index_query
Index of the query (source) point.
index_t index_match
Index of the matching (target) point.