Point Cloud Library (PCL) 1.12.1
gasd.hpp
1/*
2 * Software License Agreement (BSD License)
3 *
4 * Point Cloud Library (PCL) - www.pointclouds.org
5 * Copyright (c) 2016-, Open Perception, Inc.
6 * Copyright (c) 2016, Voxar Labs, CIn-UFPE / DEINFO-UFRPE
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
12 * are met:
13 *
14 * * Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * * Redistributions in binary form must reproduce the above
17 * copyright notice, this list of conditions and the following
18 * disclaimer in the documentation and/or other materials provided
19 * with the distribution.
20 * * Neither the name of the copyright holder(s) nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
27 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
28 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
29 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
30 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
31 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
32 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
34 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 *
37 */
38
39#ifndef PCL_FEATURES_IMPL_GASD_H_
40#define PCL_FEATURES_IMPL_GASD_H_
41
42#include <pcl/features/gasd.h>
43#include <pcl/common/common.h> // for getMinMax3D
44#include <pcl/common/transforms.h>
45
46#include <vector>
47
48//////////////////////////////////////////////////////////////////////////////////////////////
49template <typename PointInT, typename PointOutT> void
51{
53 {
54 output.width = output.height = 0;
55 output.clear ();
56 return;
57 }
58
59 // Resize the output dataset
60 output.resize (1);
61
62 // Copy header and is_dense flag from input
63 output.header = surface_->header;
64 output.is_dense = surface_->is_dense;
65
66 // Perform the actual feature computation
67 computeFeature (output);
68
70}
71
72//////////////////////////////////////////////////////////////////////////////////////////////
73template <typename PointInT, typename PointOutT> void
75{
76 Eigen::Vector4f centroid;
77 Eigen::Matrix3f covariance_matrix;
78
79 // compute centroid of the object's partial view
80 pcl::compute3DCentroid (*surface_, *indices_, centroid);
81
82 // compute covariance matrix from points and centroid of the object's partial view
83 pcl::computeCovarianceMatrix (*surface_, *indices_, centroid, covariance_matrix);
84
85 Eigen::Matrix3f eigenvectors;
86 Eigen::Vector3f eigenvalues;
87
88 // compute eigenvalues and eigenvectors of the covariance matrix
89 pcl::eigen33 (covariance_matrix, eigenvectors, eigenvalues);
90
91 // z axis of the reference frame is the eigenvector associated with the minimal eigenvalue
92 Eigen::Vector3f z_axis = eigenvectors.col (0);
93
94 // if angle between z axis and viewing direction is in the [-90 deg, 90 deg] range, then z axis is negated
95 if (z_axis.dot (view_direction_) > 0)
96 {
97 z_axis = -z_axis;
98 }
99
100 // x axis of the reference frame is the eigenvector associated with the maximal eigenvalue
101 const Eigen::Vector3f x_axis = eigenvectors.col (2);
102
103 // y axis is the cross product of z axis and x axis
104 const Eigen::Vector3f y_axis = z_axis.cross (x_axis);
105
106 const Eigen::Vector3f centroid_xyz = centroid.head<3> ();
107
108 // compute alignment transform from axes and centroid
109 transform_ << x_axis.transpose (), -x_axis.dot (centroid_xyz),
110 y_axis.transpose (), -y_axis.dot (centroid_xyz),
111 z_axis.transpose (), -z_axis.dot (centroid_xyz),
112 0.0f, 0.0f, 0.0f, 1.0f;
113}
114
115//////////////////////////////////////////////////////////////////////////////////////////////
116template <typename PointInT, typename PointOutT> void
118 const float max_coord,
119 const std::size_t half_grid_size,
120 const HistogramInterpolationMethod interp,
121 const float hbin,
122 const float hist_incr,
123 std::vector<Eigen::VectorXf> &hists)
124{
125 const std::size_t grid_size = half_grid_size * 2;
126
127 // compute normalized coordinates with respect to axis-aligned bounding cube centered on the origin
128 const Eigen::Vector3f scaled ( (p[0] / max_coord) * half_grid_size, (p[1] / max_coord) * half_grid_size, (p[2] / max_coord) * half_grid_size);
129
130 // compute histograms array coords
131 Eigen::Vector4f coords (scaled[0] + half_grid_size, scaled[1] + half_grid_size, scaled[2] + half_grid_size, hbin);
132
133 // if using histogram interpolation, subtract 0.5 so samples with the central value of the bin have full weight in it
134 if (interp != INTERP_NONE)
135 {
136 coords -= Eigen::Vector4f (0.5f, 0.5f, 0.5f, 0.5f);
137 }
138
139 // compute histograms bins indices
140 const Eigen::Vector4f bins (std::floor (coords[0]), std::floor (coords[1]), std::floor (coords[2]), std::floor (coords[3]));
141
142 // compute indices of the bin where the sample falls into
143 const std::size_t grid_idx = ( (bins[0] + 1) * (grid_size + 2) + bins[1] + 1) * (grid_size + 2) + bins[2] + 1;
144 const std::size_t h_idx = bins[3] + 1;
145
146 if (interp == INTERP_NONE)
147 {
148 // no interpolation
149 hists[grid_idx][h_idx] += hist_incr;
150 }
151 else
152 {
153 // if using histogram interpolation, compute trilinear interpolation
154 coords -= Eigen::Vector4f (bins[0], bins[1], bins[2], 0.0f);
155
156 const float v_x1 = hist_incr * coords[0];
157 const float v_x0 = hist_incr - v_x1;
158
159 const float v_xy11 = v_x1 * coords[1];
160 const float v_xy10 = v_x1 - v_xy11;
161 const float v_xy01 = v_x0 * coords[1];
162 const float v_xy00 = v_x0 - v_xy01;
163
164 const float v_xyz111 = v_xy11 * coords[2];
165 const float v_xyz110 = v_xy11 - v_xyz111;
166 const float v_xyz101 = v_xy10 * coords[2];
167 const float v_xyz100 = v_xy10 - v_xyz101;
168 const float v_xyz011 = v_xy01 * coords[2];
169 const float v_xyz010 = v_xy01 - v_xyz011;
170 const float v_xyz001 = v_xy00 * coords[2];
171 const float v_xyz000 = v_xy00 - v_xyz001;
172
173 if (interp == INTERP_TRILINEAR)
174 {
175 // trilinear interpolation
176 hists[grid_idx][h_idx] += v_xyz000;
177 hists[grid_idx + 1][h_idx] += v_xyz001;
178 hists[grid_idx + (grid_size + 2)][h_idx] += v_xyz010;
179 hists[grid_idx + (grid_size + 3)][h_idx] += v_xyz011;
180 hists[grid_idx + (grid_size + 2) * (grid_size + 2)][h_idx] += v_xyz100;
181 hists[grid_idx + (grid_size + 2) * (grid_size + 2) + 1][h_idx] += v_xyz101;
182 hists[grid_idx + (grid_size + 3) * (grid_size + 2)][h_idx] += v_xyz110;
183 hists[grid_idx + (grid_size + 3) * (grid_size + 2) + 1][h_idx] += v_xyz111;
184 }
185 else
186 {
187 // quadrilinear interpolation
188 coords[3] -= bins[3];
189
190 const float v_xyzh1111 = v_xyz111 * coords[3];
191 const float v_xyzh1110 = v_xyz111 - v_xyzh1111;
192 const float v_xyzh1101 = v_xyz110 * coords[3];
193 const float v_xyzh1100 = v_xyz110 - v_xyzh1101;
194 const float v_xyzh1011 = v_xyz101 * coords[3];
195 const float v_xyzh1010 = v_xyz101 - v_xyzh1011;
196 const float v_xyzh1001 = v_xyz100 * coords[3];
197 const float v_xyzh1000 = v_xyz100 - v_xyzh1001;
198 const float v_xyzh0111 = v_xyz011 * coords[3];
199 const float v_xyzh0110 = v_xyz011 - v_xyzh0111;
200 const float v_xyzh0101 = v_xyz010 * coords[3];
201 const float v_xyzh0100 = v_xyz010 - v_xyzh0101;
202 const float v_xyzh0011 = v_xyz001 * coords[3];
203 const float v_xyzh0010 = v_xyz001 - v_xyzh0011;
204 const float v_xyzh0001 = v_xyz000 * coords[3];
205 const float v_xyzh0000 = v_xyz000 - v_xyzh0001;
206
207 hists[grid_idx][h_idx] += v_xyzh0000;
208 hists[grid_idx][h_idx + 1] += v_xyzh0001;
209 hists[grid_idx + 1][h_idx] += v_xyzh0010;
210 hists[grid_idx + 1][h_idx + 1] += v_xyzh0011;
211 hists[grid_idx + (grid_size + 2)][h_idx] += v_xyzh0100;
212 hists[grid_idx + (grid_size + 2)][h_idx + 1] += v_xyzh0101;
213 hists[grid_idx + (grid_size + 3)][h_idx] += v_xyzh0110;
214 hists[grid_idx + (grid_size + 3)][h_idx + 1] += v_xyzh0111;
215 hists[grid_idx + (grid_size + 2) * (grid_size + 2)][h_idx] += v_xyzh1000;
216 hists[grid_idx + (grid_size + 2) * (grid_size + 2)][h_idx + 1] += v_xyzh1001;
217 hists[grid_idx + (grid_size + 2) * (grid_size + 2) + 1][h_idx] += v_xyzh1010;
218 hists[grid_idx + (grid_size + 2) * (grid_size + 2) + 1][h_idx + 1] += v_xyzh1011;
219 hists[grid_idx + (grid_size + 3) * (grid_size + 2)][h_idx] += v_xyzh1100;
220 hists[grid_idx + (grid_size + 3) * (grid_size + 2)][h_idx + 1] += v_xyzh1101;
221 hists[grid_idx + (grid_size + 3) * (grid_size + 2) + 1][h_idx] += v_xyzh1110;
222 hists[grid_idx + (grid_size + 3) * (grid_size + 2) + 1][h_idx + 1] += v_xyzh1111;
223 }
224 }
225}
226
227//////////////////////////////////////////////////////////////////////////////////////////////
228template <typename PointInT, typename PointOutT> void
230 const std::size_t hists_size,
231 const std::vector<Eigen::VectorXf> &hists,
232 PointCloudOut &output,
233 std::size_t &pos)
234{
235 for (std::size_t i = 0; i < grid_size; ++i)
236 {
237 for (std::size_t j = 0; j < grid_size; ++j)
238 {
239 for (std::size_t k = 0; k < grid_size; ++k)
240 {
241 const std::size_t idx = ( (i + 1) * (grid_size + 2) + (j + 1)) * (grid_size + 2) + (k + 1);
242
243 std::copy (hists[idx].data () + 1, hists[idx].data () + hists_size + 1, output[0].histogram + pos);
244 pos += hists_size;
245 }
246 }
247 }
248}
249
250//////////////////////////////////////////////////////////////////////////////////////////////
251template <typename PointInT, typename PointOutT> void
253{
254 // compute alignment transform using reference frame
255 computeAlignmentTransform ();
256
257 // align point cloud
258 pcl::transformPointCloud (*surface_, *indices_, shape_samples_, transform_);
259
260 const std::size_t shape_grid_size = shape_half_grid_size_ * 2;
261
262 // each histogram dimension has 2 additional bins, 1 in each boundary, for performing interpolation
263 std::vector<Eigen::VectorXf> shape_hists ((shape_grid_size + 2) * (shape_grid_size + 2) * (shape_grid_size + 2),
264 Eigen::VectorXf::Zero (shape_hists_size_ + 2));
265
266 Eigen::Vector4f centroid_p = Eigen::Vector4f::Zero ();
267
268 // compute normalization factor for distances between samples and centroid
269 Eigen::Vector4f far_pt;
270 pcl::getMaxDistance (shape_samples_, centroid_p, far_pt);
271 far_pt[3] = 0;
272 const float distance_normalization_factor = (centroid_p - far_pt).norm ();
273
274 // compute normalization factor with respect to axis-aligned bounding cube centered on the origin
275 Eigen::Vector4f min_pt, max_pt;
276 pcl::getMinMax3D (shape_samples_, min_pt, max_pt);
277
278 max_coord_ = std::max (min_pt.head<3> ().cwiseAbs ().maxCoeff (), max_pt.head<3> ().cwiseAbs ().maxCoeff ());
279
280 // normalize sample contribution with respect to the total number of points in the cloud
281 hist_incr_ = 100.0f / static_cast<float> (shape_samples_.size () - 1);
282
283 // for each sample
284 for (const auto& sample: shape_samples_)
285 {
286 // compute shape histogram array coord based on distance between sample and centroid
287 const Eigen::Vector4f p (sample.x, sample.y, sample.z, 0.0f);
288 const float d = p.norm ();
289
290 const float shape_grid_step = distance_normalization_factor / shape_half_grid_size_;
291
292 float integral;
293 const float dist_hist_val = std::modf(d / shape_grid_step, &integral);
294
295 const float dbin = dist_hist_val * shape_hists_size_;
296
297 // add sample to shape histograms, optionally performing interpolation
298 addSampleToHistograms (p, max_coord_, shape_half_grid_size_, shape_interp_, dbin, hist_incr_, shape_hists);
299 }
300
301 pos_ = 0;
302
303 // copy shape histograms to output
304 copyShapeHistogramsToOutput (shape_grid_size, shape_hists_size_, shape_hists, output, pos_);
305
306 // set remaining values of the descriptor to zero (if any)
307 std::fill (output[0].histogram + pos_, output[0].histogram + output[0].descriptorSize (), 0.0f);
308}
309
310//////////////////////////////////////////////////////////////////////////////////////////////
311template <typename PointInT, typename PointOutT> void
313 const std::size_t hists_size,
314 std::vector<Eigen::VectorXf> &hists,
315 PointCloudOut &output,
316 std::size_t &pos)
317{
318 for (std::size_t i = 0; i < grid_size; ++i)
319 {
320 for (std::size_t j = 0; j < grid_size; ++j)
321 {
322 for (std::size_t k = 0; k < grid_size; ++k)
323 {
324 const std::size_t idx = ( (i + 1) * (grid_size + 2) + (j + 1)) * (grid_size + 2) + (k + 1);
325
326 hists[idx][1] += hists[idx][hists_size + 1];
327 hists[idx][hists_size] += hists[idx][0];
328
329 std::copy (hists[idx].data () + 1, hists[idx].data () + hists_size + 1, output[0].histogram + pos);
330 pos += hists_size;
331 }
332 }
333 }
334}
335
336//////////////////////////////////////////////////////////////////////////////////////////////
337template <typename PointInT, typename PointOutT> void
339{
340 // call shape feature computation
341 GASDEstimation<PointInT, PointOutT>::computeFeature (output);
342
343 const std::size_t color_grid_size = color_half_grid_size_ * 2;
344
345 // each histogram dimension has 2 additional bins, 1 in each boundary, for performing interpolation
346 std::vector<Eigen::VectorXf> color_hists ((color_grid_size + 2) * (color_grid_size + 2) * (color_grid_size + 2),
347 Eigen::VectorXf::Zero (color_hists_size_ + 2));
348
349 // for each sample
350 for (const auto& sample: shape_samples_)
351 {
352 // compute shape histogram array coord based on distance between sample and centroid
353 const Eigen::Vector4f p (sample.x, sample.y, sample.z, 0.0f);
354
355 // compute hue value
356 float hue = 0.f;
357
358 const unsigned char max = std::max (sample.r, std::max (sample.g, sample.b));
359 const unsigned char min = std::min (sample.r, std::min (sample.g, sample.b));
360
361 const float diff_inv = 1.f / static_cast <float> (max - min);
362
363 if (std::isfinite (diff_inv))
364 {
365 if (max == sample.r)
366 {
367 hue = 60.f * (static_cast <float> (sample.g - sample.b) * diff_inv);
368 }
369 else if (max == sample.g)
370 {
371 hue = 60.f * (2.f + static_cast <float> (sample.b - sample.r) * diff_inv);
372 }
373 else
374 {
375 hue = 60.f * (4.f + static_cast <float> (sample.r - sample.g) * diff_inv); // max == b
376 }
377
378 if (hue < 0.f)
379 {
380 hue += 360.f;
381 }
382 }
383
384 // compute color histogram array coord based on hue value
385 const float hbin = (hue / 360) * color_hists_size_;
386
387 // add sample to color histograms, optionally performing interpolation
388 GASDEstimation<PointInT, PointOutT>::addSampleToHistograms (p, max_coord_, color_half_grid_size_, color_interp_, hbin, hist_incr_, color_hists);
389 }
390
391 // copy color histograms to output
392 copyColorHistogramsToOutput (color_grid_size, color_hists_size_, color_hists, output, pos_);
393
394 // set remaining values of the descriptor to zero (if any)
395 std::fill (output[0].histogram + pos_, output[0].histogram + output[0].descriptorSize (), 0.0f);
396}
397
398#define PCL_INSTANTIATE_GASDEstimation(InT, OutT) template class PCL_EXPORTS pcl::GASDEstimation<InT, OutT>;
399#define PCL_INSTANTIATE_GASDColorEstimation(InT, OutT) template class PCL_EXPORTS pcl::GASDColorEstimation<InT, OutT>;
400
401#endif // PCL_FEATURES_IMPL_GASD_H_
Feature represents the base feature class.
Definition: feature.h:107
GASDColorEstimation estimates the Globally Aligned Spatial Distribution (GASD) descriptor for a given...
Definition: gasd.h:257
GASDEstimation estimates the Globally Aligned Spatial Distribution (GASD) descriptor for a given poin...
Definition: gasd.h:75
void computeFeature(PointCloudOut &output) override
Estimate GASD descriptor.
Definition: gasd.hpp:252
void compute(PointCloudOut &output)
Overloaded computed method from pcl::Feature.
Definition: gasd.hpp:50
void addSampleToHistograms(const Eigen::Vector4f &p, const float max_coord, const std::size_t half_grid_size, const HistogramInterpolationMethod interp, const float hbin, const float hist_incr, std::vector< Eigen::VectorXf > &hists)
add a sample to its respective histogram, optionally performing interpolation.
Definition: gasd.hpp:117
Define standard C methods and C++ classes that are common to all methods.
void getMaxDistance(const pcl::PointCloud< PointT > &cloud, const Eigen::Vector4f &pivot_pt, Eigen::Vector4f &max_pt)
Get the point at maximum distance from a given point and a given pointcloud.
Definition: common.hpp:197
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
unsigned int computeCovarianceMatrix(const pcl::PointCloud< PointT > &cloud, const Eigen::Matrix< Scalar, 4, 1 > &centroid, Eigen::Matrix< Scalar, 3, 3 > &covariance_matrix)
Compute the 3x3 covariance matrix of a given set of points.
Definition: centroid.hpp:180
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.
Definition: transforms.hpp:221
void eigen33(const Matrix &mat, typename Matrix::Scalar &eigenvalue, Vector &eigenvector)
determines the eigenvector and eigenvalue of the smallest eigenvalue of the symmetric positive semi d...
Definition: eigen.hpp:296
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:56
HistogramInterpolationMethod
Different histogram interpolation methods.
Definition: gasd.h:47
@ INTERP_NONE
no interpolation
Definition: gasd.h:48