Point Cloud Library (PCL) 1.14.0
Loading...
Searching...
No Matches
ia_fpcs.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 * Copyright (C) 2008 Ben Gurion University of the Negev, Beer Sheva, Israel.
7 *
8 * All rights reserved
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions 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_REGISTRATION_IMPL_IA_FPCS_H_
39#define PCL_REGISTRATION_IMPL_IA_FPCS_H_
40
42#include <pcl/common/time.h>
43#include <pcl/common/utils.h>
44#include <pcl/registration/ia_fpcs.h>
45#include <pcl/registration/transformation_estimation_3point.h>
46#include <pcl/sample_consensus/sac_model_plane.h>
47
48#include <limits>
49
50///////////////////////////////////////////////////////////////////////////////////////////
51template <typename PointT>
52inline float
54 float max_dist,
55 int nr_threads)
56{
57 const float max_dist_sqr = max_dist * max_dist;
58 const std::size_t s = cloud->size();
59
61 tree.setInputCloud(cloud);
62
63 float mean_dist = 0.f;
64 int num = 0;
65 pcl::Indices ids(2);
66 std::vector<float> dists_sqr(2);
67
68 pcl::utils::ignore(nr_threads);
69#pragma omp parallel for default(none) shared(tree, cloud) \
70 firstprivate(ids, dists_sqr) reduction(+ : mean_dist, num) \
71 firstprivate(s, max_dist_sqr) num_threads(nr_threads)
72 for (int i = 0; i < 1000; i++) {
73 tree.nearestKSearch((*cloud)[rand() % s], 2, ids, dists_sqr);
74 if (dists_sqr[1] < max_dist_sqr) {
75 mean_dist += std::sqrt(dists_sqr[1]);
76 num++;
77 }
78 }
79
80 return (mean_dist / num);
81};
82
83///////////////////////////////////////////////////////////////////////////////////////////
84template <typename PointT>
85inline float
87 const pcl::Indices& indices,
88 float max_dist,
89 int nr_threads)
90{
91 const float max_dist_sqr = max_dist * max_dist;
92 const std::size_t s = indices.size();
93
95 tree.setInputCloud(cloud);
96
97 float mean_dist = 0.f;
98 int num = 0;
99 pcl::Indices ids(2);
100 std::vector<float> dists_sqr(2);
101
102 pcl::utils::ignore(nr_threads);
103#if OPENMP_LEGACY_CONST_DATA_SHARING_RULE
104#pragma omp parallel for default(none) shared(tree, cloud, indices) \
105 firstprivate(ids, dists_sqr) reduction(+ : mean_dist, num) num_threads(nr_threads)
106#else
107#pragma omp parallel for default(none) shared(tree, cloud, indices, s, max_dist_sqr) \
108 firstprivate(ids, dists_sqr) reduction(+ : mean_dist, num) num_threads(nr_threads)
109#endif
110 for (int i = 0; i < 1000; i++) {
111 tree.nearestKSearch((*cloud)[indices[rand() % s]], 2, ids, dists_sqr);
112 if (dists_sqr[1] < max_dist_sqr) {
113 mean_dist += std::sqrt(dists_sqr[1]);
114 num++;
115 }
116 }
117
118 return (mean_dist / num);
119};
120
121///////////////////////////////////////////////////////////////////////////////////////////
122template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
125: source_normals_()
126, target_normals_()
127, score_threshold_(std::numeric_limits<float>::max())
128, fitness_score_(std::numeric_limits<float>::max())
129{
130 reg_name_ = "pcl::registration::FPCSInitialAlignment";
131 max_iterations_ = 0;
132 ransac_iterations_ = 1000;
135}
136
137///////////////////////////////////////////////////////////////////////////////////////////
138template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
139void
141 computeTransformation(PointCloudSource& output, const Eigen::Matrix4f& guess)
142{
143 if (!initCompute())
144 return;
145
146 final_transformation_ = guess;
147 bool abort = false;
148 std::vector<MatchingCandidates> all_candidates(max_iterations_);
149 pcl::StopWatch timer;
150
151#pragma omp parallel default(none) shared(abort, all_candidates, timer) \
152 num_threads(nr_threads_)
153 {
154#ifdef _OPENMP
155 const unsigned int seed =
156 static_cast<unsigned int>(std::time(nullptr)) ^ omp_get_thread_num();
157 std::srand(seed);
158 PCL_DEBUG("[%s::computeTransformation] Using seed=%u\n", reg_name_.c_str(), seed);
159#pragma omp for schedule(dynamic)
160#endif
161 for (int i = 0; i < max_iterations_; i++) {
162#pragma omp flush(abort)
163
164 MatchingCandidates candidates(1);
165 pcl::Indices base_indices(4);
166 all_candidates[i] = candidates;
167
168 if (!abort) {
169 float ratio[2];
170 // select four coplanar point base
171 if (selectBase(base_indices, ratio) == 0) {
172 // calculate candidate pair correspondences using diagonal lengths of base
173 pcl::Correspondences pairs_a, pairs_b;
174 if (bruteForceCorrespondences(base_indices[0], base_indices[1], pairs_a) ==
175 0 &&
176 bruteForceCorrespondences(base_indices[2], base_indices[3], pairs_b) ==
177 0) {
178 // determine candidate matches by combining pair correspondences based on
179 // segment distances
180 std::vector<pcl::Indices> matches;
181 if (determineBaseMatches(base_indices, matches, pairs_a, pairs_b, ratio) ==
182 0) {
183 // check and evaluate candidate matches and store them
184 handleMatches(base_indices, matches, candidates);
185 if (!candidates.empty())
186 all_candidates[i] = candidates;
187 }
188 }
189 }
190
191 // check terminate early (time or fitness_score threshold reached)
192 abort = (!candidates.empty() ? candidates[0].fitness_score < score_threshold_
193 : abort);
194 abort = (abort ? abort : timer.getTimeSeconds() > max_runtime_);
195
196#pragma omp flush(abort)
197 }
198 }
199 }
200
201 // determine best match over all tries
202 finalCompute(all_candidates);
203
204 // apply the final transformation
205 pcl::transformPointCloud(*input_, output, final_transformation_);
206
207 deinitCompute();
208}
209
210///////////////////////////////////////////////////////////////////////////////////////////
211template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
212bool
215{
216 const unsigned int seed = std::time(nullptr);
217 std::srand(seed);
218 PCL_DEBUG("[%s::initCompute] Using seed=%u\n", reg_name_.c_str(), seed);
219
220 // basic pcl initialization
222 return (false);
223
224 // check if source and target are given
225 if (!input_ || !target_) {
226 PCL_ERROR("[%s::initCompute] Source or target dataset not given!\n",
227 reg_name_.c_str());
228 return (false);
229 }
230
231 if (!target_indices_ || target_indices_->empty()) {
232 target_indices_.reset(new pcl::Indices(target_->size()));
233 int index = 0;
234 for (auto& target_index : *target_indices_)
235 target_index = index++;
236 target_cloud_updated_ = true;
237 }
238
239 // if a sample size for the point clouds is given; preferably no sampling of target
240 // cloud
241 if (nr_samples_ != 0) {
242 const int ss = static_cast<int>(indices_->size());
243 const int sample_fraction_src = std::max(1, static_cast<int>(ss / nr_samples_));
244
245 source_indices_ = pcl::IndicesPtr(new pcl::Indices);
246 for (int i = 0; i < ss; i++)
247 if (rand() % sample_fraction_src == 0)
248 source_indices_->push_back((*indices_)[i]);
249 }
250 else
251 source_indices_ = indices_;
252
253 // check usage of normals
254 if (source_normals_ && target_normals_ && source_normals_->size() == input_->size() &&
255 target_normals_->size() == target_->size())
256 use_normals_ = true;
257
258 // set up tree structures
259 if (target_cloud_updated_) {
260 tree_->setInputCloud(target_, target_indices_);
261 target_cloud_updated_ = false;
262 }
263
264 // set predefined variables
265 constexpr int min_iterations = 4;
266 constexpr float diameter_fraction = 0.3f;
267
268 // get diameter of input cloud (distance between farthest points)
269 Eigen::Vector4f pt_min, pt_max;
270 pcl::getMinMax3D(*target_, *target_indices_, pt_min, pt_max);
271 diameter_ = (pt_max - pt_min).norm();
272
273 // derive the limits for the random base selection
274 float max_base_diameter = diameter_ * approx_overlap_ * 2.f;
275 max_base_diameter_sqr_ = max_base_diameter * max_base_diameter;
276
277 // normalize the delta
278 if (normalize_delta_) {
279 float mean_dist = getMeanPointDensity<PointTarget>(
280 target_, *target_indices_, 0.05f * diameter_, nr_threads_);
281 delta_ *= mean_dist;
282 }
283
284 // heuristic determination of number of trials to have high probability of finding a
285 // good solution
286 if (max_iterations_ == 0) {
287 float first_est = std::log(small_error_) /
288 std::log(1.0 - std::pow(static_cast<double>(approx_overlap_),
289 static_cast<double>(min_iterations)));
290 max_iterations_ =
291 static_cast<int>(first_est / (diameter_fraction * approx_overlap_ * 2.f));
292 }
293
294 // set further parameter
295 if (score_threshold_ == std::numeric_limits<float>::max())
296 score_threshold_ = 1.f - approx_overlap_;
297
298 if (max_iterations_ < 4)
299 max_iterations_ = 4;
300
301 if (max_runtime_ < 1)
302 max_runtime_ = std::numeric_limits<int>::max();
303
304 // calculate internal parameters based on the the estimated point density
305 max_pair_diff_ = delta_ * 2.f;
306 max_edge_diff_ = delta_ * 4.f;
307 coincidation_limit_ = delta_ * 2.f; // EDITED: originally std::sqrt (delta_ * 2.f)
308 max_mse_ = powf(delta_ * 2.f, 2.f);
309 max_inlier_dist_sqr_ = powf(delta_ * 2.f, 2.f);
310
311 // reset fitness_score
312 fitness_score_ = std::numeric_limits<float>::max();
313
314 return (true);
315}
316
317///////////////////////////////////////////////////////////////////////////////////////////
318template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
319int
321 selectBase(pcl::Indices& base_indices, float (&ratio)[2])
322{
323 const float too_close_sqr = max_base_diameter_sqr_ * 0.01;
324
325 Eigen::VectorXf coefficients(4);
327 plane.setIndices(target_indices_);
328 Eigen::Vector4f centre_pt;
329 float nearest_to_plane = std::numeric_limits<float>::max();
330
331 // repeat base search until valid quadruple was found or ransac_iterations_ number of
332 // tries were unsuccessful
333 for (int i = 0; i < ransac_iterations_; i++) {
334 // random select an appropriate point triple
335 if (selectBaseTriangle(base_indices) < 0)
336 continue;
337
338 pcl::Indices base_triple(base_indices.begin(), base_indices.end() - 1);
339 plane.computeModelCoefficients(base_triple, coefficients);
340 pcl::compute3DCentroid(*target_, base_triple, centre_pt);
341
342 // loop over all points in source cloud to find most suitable fourth point
343 const PointTarget* pt1 = &((*target_)[base_indices[0]]);
344 const PointTarget* pt2 = &((*target_)[base_indices[1]]);
345 const PointTarget* pt3 = &((*target_)[base_indices[2]]);
346
347 for (const auto& target_index : *target_indices_) {
348 const PointTarget* pt4 = &((*target_)[target_index]);
350 float d1 = pcl::squaredEuclideanDistance(*pt4, *pt1);
351 float d2 = pcl::squaredEuclideanDistance(*pt4, *pt2);
352 float d3 = pcl::squaredEuclideanDistance(*pt4, *pt3);
353 float d4 = (pt4->getVector3fMap() - centre_pt.head(3)).squaredNorm();
354
355 // check distance between points w.r.t minimum sampling distance; EDITED -> 4th
356 // point now also limited by max base line
357 if (d1 < too_close_sqr || d2 < too_close_sqr || d3 < too_close_sqr ||
358 d4 < too_close_sqr || d1 > max_base_diameter_sqr_ ||
359 d2 > max_base_diameter_sqr_ || d3 > max_base_diameter_sqr_)
360 continue;
361
362 // check distance to plane to get point closest to plane
363 float dist_to_plane = pcl::pointToPlaneDistance(*pt4, coefficients);
364 if (dist_to_plane < nearest_to_plane) {
365 base_indices[3] = target_index;
366 nearest_to_plane = dist_to_plane;
367 }
368 }
369
370 // check if at least one point fulfilled the conditions
371 if (nearest_to_plane != std::numeric_limits<float>::max()) {
372 // order points to build largest quadrangle and calculate intersection ratios of
373 // diagonals
374 setupBase(base_indices, ratio);
375 return (0);
376 }
377 }
379 // return unsuccessful if no quadruple was selected
380 return (-1);
381}
382
383///////////////////////////////////////////////////////////////////////////////////////////
384template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
385int
388{
389 const auto nr_points = target_indices_->size();
390 float best_t = 0.f;
391
392 // choose random first point
393 base_indices[0] = (*target_indices_)[rand() % nr_points];
394 auto* index1 = base_indices.data();
395
396 // random search for 2 other points (as far away as overlap allows)
397 for (int i = 0; i < ransac_iterations_; i++) {
398 auto* index2 = &(*target_indices_)[rand() % nr_points];
399 auto* index3 = &(*target_indices_)[rand() % nr_points];
400
401 Eigen::Vector3f u =
402 (*target_)[*index2].getVector3fMap() - (*target_)[*index1].getVector3fMap();
403 Eigen::Vector3f v =
404 (*target_)[*index3].getVector3fMap() - (*target_)[*index1].getVector3fMap();
405 float t =
406 u.cross(v).squaredNorm(); // triangle area (0.5 * sqrt(t)) should be maximal
408 // check for most suitable point triple
409 if (t > best_t && u.squaredNorm() < max_base_diameter_sqr_ &&
410 v.squaredNorm() < max_base_diameter_sqr_) {
411 best_t = t;
412 base_indices[1] = *index2;
413 base_indices[2] = *index3;
414 }
415 }
416
417 // return if a triplet could be selected
418 return (best_t == 0.f ? -1 : 0);
420
421///////////////////////////////////////////////////////////////////////////////////////////
422template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
423void
425 setupBase(pcl::Indices& base_indices, float (&ratio)[2])
426{
427 float best_t = std::numeric_limits<float>::max();
428 const pcl::Indices copy(base_indices.begin(), base_indices.end());
429 pcl::Indices temp(base_indices.begin(), base_indices.end());
430
431 // loop over all combinations of base points
432 for (auto i = copy.begin(), i_e = copy.end(); i != i_e; ++i)
433 for (auto j = copy.begin(), j_e = copy.end(); j != j_e; ++j) {
434 if (i == j)
435 continue;
436
437 for (auto k = copy.begin(), k_e = copy.end(); k != k_e; ++k) {
438 if (k == j || k == i)
439 continue;
440
441 auto l = copy.begin();
442 while (l == i || l == j || l == k)
443 ++l;
444
445 temp[0] = *i;
446 temp[1] = *j;
447 temp[2] = *k;
448 temp[3] = *l;
449
450 // calculate diagonal intersection ratios and check for suitable segment to
451 // segment distances
452 float ratio_temp[2];
453 float t = segmentToSegmentDist(temp, ratio_temp);
454 if (t < best_t) {
455 best_t = t;
456 ratio[0] = ratio_temp[0];
457 ratio[1] = ratio_temp[1];
458 base_indices = temp;
459 }
460 }
461 }
463
464///////////////////////////////////////////////////////////////////////////////////////////
465template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
466float
468 segmentToSegmentDist(const pcl::Indices& base_indices, float (&ratio)[2])
469{
470 // get point vectors
471 Eigen::Vector3f u = (*target_)[base_indices[1]].getVector3fMap() -
472 (*target_)[base_indices[0]].getVector3fMap();
473 Eigen::Vector3f v = (*target_)[base_indices[3]].getVector3fMap() -
474 (*target_)[base_indices[2]].getVector3fMap();
475 Eigen::Vector3f w = (*target_)[base_indices[0]].getVector3fMap() -
476 (*target_)[base_indices[2]].getVector3fMap();
477
478 // calculate segment distances
479 float a = u.dot(u);
480 float b = u.dot(v);
481 float c = v.dot(v);
482 float d = u.dot(w);
483 float e = v.dot(w);
484 float D = a * c - b * b;
485 float sN = 0.f, sD = D;
486 float tN = 0.f, tD = D;
487
488 // check segments
489 if (D < small_error_) {
490 sN = 0.f;
491 sD = 1.f;
492 tN = e;
493 tD = c;
494 }
495 else {
496 sN = (b * e - c * d);
497 tN = (a * e - b * d);
498
499 if (sN < 0.f) {
500 sN = 0.f;
501 tN = e;
502 tD = c;
503 }
504 else if (sN > sD) {
505 sN = sD;
506 tN = e + b;
507 tD = c;
508 }
509 }
510
511 if (tN < 0.f) {
512 tN = 0.f;
513
514 if (-d < 0.f)
515 sN = 0.f;
516
517 else if (-d > a)
518 sN = sD;
519
520 else {
521 sN = -d;
522 sD = a;
523 }
524 }
525
526 else if (tN > tD) {
527 tN = tD;
528
529 if ((-d + b) < 0.f)
530 sN = 0.f;
531
532 else if ((-d + b) > a)
533 sN = sD;
534
535 else {
536 sN = (-d + b);
537 sD = a;
538 }
539 }
540
541 // set intersection ratios
542 ratio[0] = (std::abs(sN) < small_error_) ? 0.f : sN / sD;
543 ratio[1] = (std::abs(tN) < small_error_) ? 0.f : tN / tD;
544
545 Eigen::Vector3f x = w + (ratio[0] * u) - (ratio[1] * v);
546 return (x.norm());
547}
548
549///////////////////////////////////////////////////////////////////////////////////////////
550template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
551int
553 bruteForceCorrespondences(int idx1, int idx2, pcl::Correspondences& pairs)
554{
555 const float max_norm_diff = 0.5f * max_norm_diff_ * M_PI / 180.f;
556
557 // calculate reference segment distance and normal angle
558 float ref_dist = pcl::euclideanDistance((*target_)[idx1], (*target_)[idx2]);
559 float ref_norm_angle =
560 (use_normals_ ? ((*target_normals_)[idx1].getNormalVector3fMap() -
561 (*target_normals_)[idx2].getNormalVector3fMap())
562 .norm()
563 : 0.f);
564
565 // loop over all pairs of points in source point cloud
566 auto it_out = source_indices_->begin(), it_out_e = source_indices_->end() - 1;
567 auto it_in_e = source_indices_->end();
568 for (; it_out != it_out_e; it_out++) {
569 auto it_in = it_out + 1;
570 const PointSource* pt1 = &(*input_)[*it_out];
571 for (; it_in != it_in_e; it_in++) {
572 const PointSource* pt2 = &(*input_)[*it_in];
573
574 // check point distance compared to reference dist (from base)
575 float dist = pcl::euclideanDistance(*pt1, *pt2);
576 if (std::abs(dist - ref_dist) < max_pair_diff_) {
577 // add here normal evaluation if normals are given
578 if (use_normals_) {
579 const NormalT* pt1_n = &((*source_normals_)[*it_out]);
580 const NormalT* pt2_n = &((*source_normals_)[*it_in]);
581
582 float norm_angle_1 =
583 (pt1_n->getNormalVector3fMap() - pt2_n->getNormalVector3fMap()).norm();
584 float norm_angle_2 =
585 (pt1_n->getNormalVector3fMap() + pt2_n->getNormalVector3fMap()).norm();
586
587 float norm_diff = std::min<float>(std::abs(norm_angle_1 - ref_norm_angle),
588 std::abs(norm_angle_2 - ref_norm_angle));
589 if (norm_diff > max_norm_diff)
590 continue;
591 }
592
593 pairs.push_back(pcl::Correspondence(*it_in, *it_out, dist));
594 pairs.push_back(pcl::Correspondence(*it_out, *it_in, dist));
595 }
596 }
597 }
598
599 // return success if at least one correspondence was found
600 return (pairs.empty() ? -1 : 0);
601}
602
603///////////////////////////////////////////////////////////////////////////////////////////
604template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
605int
607 determineBaseMatches(const pcl::Indices& base_indices,
608 std::vector<pcl::Indices>& matches,
609 const pcl::Correspondences& pairs_a,
610 const pcl::Correspondences& pairs_b,
611 const float (&ratio)[2])
612{
613 // calculate edge lengths of base
614 float dist_base[4];
615 dist_base[0] =
616 pcl::euclideanDistance((*target_)[base_indices[0]], (*target_)[base_indices[2]]);
617 dist_base[1] =
618 pcl::euclideanDistance((*target_)[base_indices[0]], (*target_)[base_indices[3]]);
619 dist_base[2] =
620 pcl::euclideanDistance((*target_)[base_indices[1]], (*target_)[base_indices[2]]);
621 dist_base[3] =
622 pcl::euclideanDistance((*target_)[base_indices[1]], (*target_)[base_indices[3]]);
623
624 // loop over first point pair correspondences and store intermediate points 'e' in new
625 // point cloud
627 cloud_e->resize(pairs_a.size() * 2);
628 auto it_pt = cloud_e->begin();
629 for (const auto& pair : pairs_a) {
630 const PointSource* pt1 = &((*input_)[pair.index_match]);
631 const PointSource* pt2 = &((*input_)[pair.index_query]);
632
633 // calculate intermediate points using both ratios from base (r1,r2)
634 for (int i = 0; i < 2; i++, it_pt++) {
635 it_pt->x = pt1->x + ratio[i] * (pt2->x - pt1->x);
636 it_pt->y = pt1->y + ratio[i] * (pt2->y - pt1->y);
637 it_pt->z = pt1->z + ratio[i] * (pt2->z - pt1->z);
638 }
639 }
640
641 // initialize new kd tree of intermediate points from first point pair correspondences
643 tree_e->setInputCloud(cloud_e);
644
645 pcl::Indices ids;
646 std::vector<float> dists_sqr;
647
648 // loop over second point pair correspondences
649 for (const auto& pair : pairs_b) {
650 const PointTarget* pt1 = &((*input_)[pair.index_match]);
651 const PointTarget* pt2 = &((*input_)[pair.index_query]);
652
653 // calculate intermediate points using both ratios from base (r1,r2)
654 for (const float& r : ratio) {
655 PointTarget pt_e;
656 pt_e.x = pt1->x + r * (pt2->x - pt1->x);
657 pt_e.y = pt1->y + r * (pt2->y - pt1->y);
658 pt_e.z = pt1->z + r * (pt2->z - pt1->z);
659
660 // search for corresponding intermediate points
661 tree_e->radiusSearch(pt_e, coincidation_limit_, ids, dists_sqr);
662 for (const auto& id : ids) {
663 pcl::Indices match_indices(4);
664
665 match_indices[0] =
666 pairs_a[static_cast<int>(std::floor((id / 2.f)))].index_match;
667 match_indices[1] =
668 pairs_a[static_cast<int>(std::floor((id / 2.f)))].index_query;
669 match_indices[2] = pair.index_match;
670 match_indices[3] = pair.index_query;
671
672 // EDITED: added coarse check of match based on edge length (due to rigid-body )
673 if (checkBaseMatch(match_indices, dist_base) < 0)
674 continue;
675
676 matches.push_back(match_indices);
677 }
678 }
679 }
680
681 // return unsuccessful if no match was found
682 return (!matches.empty() ? 0 : -1);
683}
684
685///////////////////////////////////////////////////////////////////////////////////////////
686template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
687int
689 checkBaseMatch(const pcl::Indices& match_indices, const float (&dist_ref)[4])
690{
691 float d0 =
692 pcl::euclideanDistance((*input_)[match_indices[0]], (*input_)[match_indices[2]]);
693 float d1 =
694 pcl::euclideanDistance((*input_)[match_indices[0]], (*input_)[match_indices[3]]);
695 float d2 =
696 pcl::euclideanDistance((*input_)[match_indices[1]], (*input_)[match_indices[2]]);
697 float d3 =
698 pcl::euclideanDistance((*input_)[match_indices[1]], (*input_)[match_indices[3]]);
699
700 // check edge distances of match w.r.t the base
701 return (std::abs(d0 - dist_ref[0]) < max_edge_diff_ &&
702 std::abs(d1 - dist_ref[1]) < max_edge_diff_ &&
703 std::abs(d2 - dist_ref[2]) < max_edge_diff_ &&
704 std::abs(d3 - dist_ref[3]) < max_edge_diff_)
705 ? 0
706 : -1;
707}
708
709///////////////////////////////////////////////////////////////////////////////////////////
710template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
711void
713 handleMatches(const pcl::Indices& base_indices,
714 std::vector<pcl::Indices>& matches,
715 MatchingCandidates& candidates)
716{
717 candidates.resize(1);
718 float fitness_score = std::numeric_limits<float>::max();
719
720 // loop over all Candidate matches
721 for (auto& match : matches) {
722 Eigen::Matrix4f transformation_temp;
723 pcl::Correspondences correspondences_temp;
724
725 // determine corresondences between base and match according to their distance to
726 // centroid
727 linkMatchWithBase(base_indices, match, correspondences_temp);
728
729 // check match based on residuals of the corresponding points after
730 if (validateMatch(base_indices, match, correspondences_temp, transformation_temp) <
731 0)
732 continue;
733
734 // check resulting using a sub sample of the source point cloud and compare to
735 // previous matches
736 if (validateTransformation(transformation_temp, fitness_score) < 0)
737 continue;
738
739 // store best match as well as associated fitness_score and transformation
740 candidates[0].fitness_score = fitness_score;
741 candidates[0].transformation = transformation_temp;
742 correspondences_temp.erase(correspondences_temp.end() - 1);
743 candidates[0].correspondences = correspondences_temp;
744 }
745}
746
747///////////////////////////////////////////////////////////////////////////////////////////
748template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
749void
751 linkMatchWithBase(const pcl::Indices& base_indices,
752 pcl::Indices& match_indices,
753 pcl::Correspondences& correspondences)
754{
755 // calculate centroid of base and target
756 Eigen::Vector4f centre_base{0, 0, 0, 0}, centre_match{0, 0, 0, 0};
757 pcl::compute3DCentroid(*target_, base_indices, centre_base);
758 pcl::compute3DCentroid(*input_, match_indices, centre_match);
759
760 PointTarget centre_pt_base;
761 centre_pt_base.x = centre_base[0];
762 centre_pt_base.y = centre_base[1];
763 centre_pt_base.z = centre_base[2];
764
765 PointSource centre_pt_match;
766 centre_pt_match.x = centre_match[0];
767 centre_pt_match.y = centre_match[1];
768 centre_pt_match.z = centre_match[2];
769
770 // find corresponding points according to their distance to the centroid
771 pcl::Indices copy = match_indices;
772
773 auto it_match_orig = match_indices.begin();
774 for (auto it_base = base_indices.cbegin(), it_base_e = base_indices.cend();
775 it_base != it_base_e;
776 it_base++, it_match_orig++) {
777 float dist_sqr_1 =
778 pcl::squaredEuclideanDistance((*target_)[*it_base], centre_pt_base);
779 float best_diff_sqr = std::numeric_limits<float>::max();
780 int best_index = -1;
781
782 for (const auto& match_index : copy) {
783 // calculate difference of distances to centre point
784 float dist_sqr_2 =
785 pcl::squaredEuclideanDistance((*input_)[match_index], centre_pt_match);
786 float diff_sqr = std::abs(dist_sqr_1 - dist_sqr_2);
787
788 if (diff_sqr < best_diff_sqr) {
789 best_diff_sqr = diff_sqr;
790 best_index = match_index;
791 }
792 }
793
794 // assign new correspondence and update indices of matched targets
795 correspondences.push_back(pcl::Correspondence(best_index, *it_base, best_diff_sqr));
796 *it_match_orig = best_index;
797 }
798}
799
800///////////////////////////////////////////////////////////////////////////////////////////
801template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
802int
804 validateMatch(const pcl::Indices& base_indices,
805 const pcl::Indices& match_indices,
806 const pcl::Correspondences& correspondences,
807 Eigen::Matrix4f& transformation)
808{
809 // only use triplet of points to simplify process (possible due to planar case)
810 pcl::Correspondences correspondences_temp = correspondences;
811 correspondences_temp.erase(correspondences_temp.end() - 1);
812
813 // estimate transformation between correspondence set
814 transformation_estimation_->estimateRigidTransformation(
815 *input_, *target_, correspondences_temp, transformation);
816
817 // transform base points
818 PointCloudSource match_transformed;
819 pcl::transformPointCloud(*input_, match_indices, match_transformed, transformation);
820
821 // calculate residuals of transformation and check against maximum threshold
822 std::size_t nr_points = correspondences_temp.size();
823 float mse = 0.f;
824 for (std::size_t i = 0; i < nr_points; i++)
825 mse += pcl::squaredEuclideanDistance(match_transformed.points[i],
826 target_->points[base_indices[i]]);
827
828 mse /= nr_points;
829 return (mse < max_mse_ ? 0 : -1);
830}
831
832///////////////////////////////////////////////////////////////////////////////////////////
833template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
834int
836 validateTransformation(Eigen::Matrix4f& transformation, float& fitness_score)
837{
838 // transform source point cloud
839 PointCloudSource source_transformed;
841 *input_, *source_indices_, source_transformed, transformation);
842
843 std::size_t nr_points = source_transformed.size();
844 std::size_t terminate_value =
845 fitness_score > 1 ? 0
846 : static_cast<std::size_t>((1.f - fitness_score) * nr_points);
847
848 float inlier_score_temp = 0;
849 pcl::Indices ids;
850 std::vector<float> dists_sqr;
851 auto it = source_transformed.begin();
852
853 for (std::size_t i = 0; i < nr_points; it++, i++) {
854 // search for nearest point using kd tree search
855 tree_->nearestKSearch(*it, 1, ids, dists_sqr);
856 inlier_score_temp += (dists_sqr[0] < max_inlier_dist_sqr_ ? 1 : 0);
857
858 // early terminating
859 if (nr_points - i + inlier_score_temp < terminate_value)
860 break;
861 }
862
863 // check current costs and return unsuccessful if larger than previous ones
864 inlier_score_temp /= static_cast<float>(nr_points);
865 float fitness_score_temp = 1.f - inlier_score_temp;
866
867 if (fitness_score_temp > fitness_score)
868 return (-1);
869
870 fitness_score = fitness_score_temp;
871 return (0);
872}
873
874///////////////////////////////////////////////////////////////////////////////////////////
875template <typename PointSource, typename PointTarget, typename NormalT, typename Scalar>
876void
878 finalCompute(const std::vector<MatchingCandidates>& candidates)
879{
880 // get best fitness_score over all tries
881 int nr_candidates = static_cast<int>(candidates.size());
882 int best_index = -1;
883 float best_score = std::numeric_limits<float>::max();
884 for (int i = 0; i < nr_candidates; i++) {
885 const float& fitness_score = candidates[i][0].fitness_score;
886 if (fitness_score < best_score) {
887 best_score = fitness_score;
888 best_index = i;
889 }
890 }
891
892 // check if a valid candidate was available
893 if (!(best_index < 0)) {
894 fitness_score_ = candidates[best_index][0].fitness_score;
895 final_transformation_ = candidates[best_index][0].transformation;
896 *correspondences_ = candidates[best_index][0].correspondences;
897
898 // here we define convergence if resulting fitness_score is below 1-threshold
899 converged_ = fitness_score_ < score_threshold_;
900 }
901}
902
903///////////////////////////////////////////////////////////////////////////////////////////
904
905#endif // PCL_REGISTRATION_IMPL_IA_4PCS_H_
PCL base class.
Definition pcl_base.h:70
std::size_t size() const
iterator begin() noexcept
std::vector< PointT, Eigen::aligned_allocator< PointT > > points
The point data.
shared_ptr< const PointCloud< PointT > > ConstPtr
typename KdTreeReciprocal::Ptr KdTreeReciprocalPtr
std::string reg_name_
The registration method name.
typename PointCloudSource::Ptr PointCloudSourcePtr
int ransac_iterations_
The number of iterations RANSAC should run for.
TransformationEstimationPtr transformation_estimation_
A TransformationEstimation object, used to calculate the 4x4 rigid transformation.
int max_iterations_
The maximum number of iterations the internal optimization should run for.
void setIndices(const IndicesPtr &indices)
Provide a pointer to the vector of indices that represents the input data.
Definition sac_model.h:324
SampleConsensusModelPlane defines a model for 3D plane segmentation.
bool computeModelCoefficients(const Indices &samples, Eigen::VectorXf &model_coefficients) const override
Check whether the given index samples can form a valid plane model, compute the model coefficients fr...
Simple stopwatch.
Definition time.h:59
double getTimeSeconds() const
Retrieve the time in seconds spent since the last call to reset().
Definition time.h:74
virtual void finalCompute(const std::vector< MatchingCandidates > &candidates)
Final computation of best match out of vector of best matches.
Definition ia_fpcs.hpp:878
void setupBase(pcl::Indices &base_indices, float(&ratio)[2])
Setup the base (four coplanar points) by ordering the points and computing intersection ratios and se...
Definition ia_fpcs.hpp:425
int selectBaseTriangle(pcl::Indices &base_indices)
Select randomly a triplet of points with large point-to-point distances.
Definition ia_fpcs.hpp:387
virtual int bruteForceCorrespondences(int idx1, int idx2, pcl::Correspondences &pairs)
Search for corresponding point pairs given the distance between two base points.
Definition ia_fpcs.hpp:553
int selectBase(pcl::Indices &base_indices, float(&ratio)[2])
Select an approximately coplanar set of four points from the source cloud.
Definition ia_fpcs.hpp:321
virtual int validateTransformation(Eigen::Matrix4f &transformation, float &fitness_score)
Validate the transformation by calculating the number of inliers after transforming the source cloud.
Definition ia_fpcs.hpp:836
virtual int determineBaseMatches(const pcl::Indices &base_indices, std::vector< pcl::Indices > &matches, const pcl::Correspondences &pairs_a, const pcl::Correspondences &pairs_b, const float(&ratio)[2])
Determine base matches by combining the point pair candidate and search for coinciding intersection p...
Definition ia_fpcs.hpp:607
virtual void linkMatchWithBase(const pcl::Indices &base_indices, pcl::Indices &match_indices, pcl::Correspondences &correspondences)
Sets the correspondences between the base B and the match M by using the distance of each point to th...
Definition ia_fpcs.hpp:751
virtual void handleMatches(const pcl::Indices &base_indices, std::vector< pcl::Indices > &matches, MatchingCandidates &candidates)
Method to handle current candidate matches.
Definition ia_fpcs.hpp:713
int checkBaseMatch(const pcl::Indices &match_indices, const float(&ds)[4])
Check if outer rectangle distance of matched points fit with the base rectangle.
Definition ia_fpcs.hpp:689
float segmentToSegmentDist(const pcl::Indices &base_indices, float(&ratio)[2])
Calculate intersection ratios and segment to segment distances of base diagonals.
Definition ia_fpcs.hpp:468
virtual int validateMatch(const pcl::Indices &base_indices, const pcl::Indices &match_indices, const pcl::Correspondences &correspondences, Eigen::Matrix4f &transformation)
Validate the matching by computing the transformation between the source and target based on the four...
Definition ia_fpcs.hpp:804
void computeTransformation(PointCloudSource &output, const Eigen::Matrix4f &guess) override
Rigid transformation computation method.
Definition ia_fpcs.hpp:141
virtual bool initCompute()
Internal computation initialization.
Definition ia_fpcs.hpp:214
TransformationEstimation3Points represents the class for transformation estimation based on:
search::KdTree is a wrapper class which inherits the pcl::KdTree class for performing search function...
Definition kdtree.h:62
int nearestKSearch(const PointT &point, int k, Indices &k_indices, std::vector< float > &k_sqr_distances) const override
Search for the k-nearest neighbors for the given query point.
Definition kdtree.hpp:88
bool setInputCloud(const PointCloudConstPtr &cloud, const IndicesConstPtr &indices=IndicesConstPtr()) override
Provide a pointer to the input dataset.
Definition kdtree.hpp:76
Define standard C methods to do distance calculations.
Define methods for measuring time spent in code blocks.
void getMinMax3D(const pcl::PointCloud< PointT > &cloud, PointT &min_pt, PointT &max_pt)
Get the minimum and maximum values on each of the 3 (x-y-z) dimensions in a given pointcloud.
Definition common.hpp:295
void transformPointCloud(const pcl::PointCloud< PointT > &cloud_in, pcl::PointCloud< PointT > &cloud_out, const Eigen::Matrix< Scalar, 4, 4 > &transform, bool copy_all_fields)
Apply a rigid transform defined by a 4x4 matrix.
unsigned int compute3DCentroid(ConstCloudIterator< PointT > &cloud_iterator, Eigen::Matrix< Scalar, 4, 1 > &centroid)
Compute the 3D (X-Y-Z) centroid of a set of points and return it as a 3D vector.
Definition centroid.hpp:57
double pointToPlaneDistance(const Point &p, double a, double b, double c, double d)
Get the distance from a point to a plane (unsigned) defined by ax+by+cz+d=0.
std::vector< MatchingCandidate, Eigen::aligned_allocator< MatchingCandidate > > MatchingCandidates
void ignore(const T &...)
Utility function to eliminate unused variable warnings.
Definition utils.h:62
float squaredEuclideanDistance(const PointType1 &p1, const PointType2 &p2)
Calculate the squared euclidean distance between the two given points.
Definition distances.h:182
float getMeanPointDensity(const typename pcl::PointCloud< PointT >::ConstPtr &cloud, float max_dist, int nr_threads=1)
Compute the mean point density of a given point cloud.
Definition ia_fpcs.hpp:53
std::vector< pcl::Correspondence, Eigen::aligned_allocator< pcl::Correspondence > > Correspondences
float euclideanDistance(const PointType1 &p1, const PointType2 &p2)
Calculate the euclidean distance between the two given points.
Definition distances.h:204
IndicesAllocator<> Indices
Type used for indices in PCL.
Definition types.h:133
shared_ptr< Indices > IndicesPtr
Definition pcl_base.h:58
#define M_PI
Definition pcl_macros.h:201
Correspondence represents a match between two entities (e.g., points, descriptors,...
A point structure representing normal coordinates and the surface curvature estimate.