Point Cloud Library (PCL)  1.7.1
spin_image.hpp
1 /*
2  * Software License Agreement (BSD License)
3  *
4  * Point Cloud Library (PCL) - www.pointclouds.org
5  * Copyright (c) 2010-2012, Willow Garage, Inc.
6  * Copyright (c) 2012-, Open Perception, Inc.
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  * $Id$
38  *
39  */
40 
41 #ifndef PCL_FEATURES_IMPL_SPIN_IMAGE_H_
42 #define PCL_FEATURES_IMPL_SPIN_IMAGE_H_
43 
44 #include <limits>
45 #include <pcl/point_cloud.h>
46 #include <pcl/point_types.h>
47 #include <pcl/exceptions.h>
48 #include <pcl/kdtree/kdtree_flann.h>
49 #include <pcl/features/spin_image.h>
50 #include <cmath>
51 
52 //////////////////////////////////////////////////////////////////////////////////////////////
53 template <typename PointInT, typename PointNT, typename PointOutT>
55  unsigned int image_width, double support_angle_cos, unsigned int min_pts_neighb) :
56  input_normals_ (), rotation_axes_cloud_ (),
57  is_angular_ (false), rotation_axis_ (), use_custom_axis_(false), use_custom_axes_cloud_ (false),
58  is_radial_ (false), image_width_ (image_width), support_angle_cos_ (support_angle_cos),
59  min_pts_neighb_ (min_pts_neighb)
60 {
61  assert (support_angle_cos_ <= 1.0 && support_angle_cos_ >= 0.0); // may be permit negative cosine?
62 
63  feature_name_ = "SpinImageEstimation";
64 }
65 
66 
67 //////////////////////////////////////////////////////////////////////////////////////////////
68 template <typename PointInT, typename PointNT, typename PointOutT> Eigen::ArrayXXd
70 {
71  assert (image_width_ > 0);
72  assert (support_angle_cos_ <= 1.0 && support_angle_cos_ >= 0.0); // may be permit negative cosine?
73 
74  const Eigen::Vector3f origin_point (input_->points[index].getVector3fMap ());
75 
76  Eigen::Vector3f origin_normal;
77  origin_normal =
78  input_normals_ ?
79  input_normals_->points[index].getNormalVector3fMap () :
80  Eigen::Vector3f (); // just a placeholder; should never be used!
81 
82  const Eigen::Vector3f rotation_axis = use_custom_axis_ ?
83  rotation_axis_.getNormalVector3fMap () :
84  use_custom_axes_cloud_ ?
85  rotation_axes_cloud_->points[index].getNormalVector3fMap () :
86  origin_normal;
87 
88  Eigen::ArrayXXd m_matrix (Eigen::ArrayXXd::Zero (image_width_+1, 2*image_width_+1));
89  Eigen::ArrayXXd m_averAngles (Eigen::ArrayXXd::Zero (image_width_+1, 2*image_width_+1));
90 
91  // OK, we are interested in the points of the cylinder of height 2*r and
92  // base radius r, where r = m_dBinSize * in_iImageWidth
93  // it can be embedded to the sphere of radius sqrt(2) * m_dBinSize * in_iImageWidth
94  // suppose that points are uniformly distributed, so we lose ~40%
95  // according to the volumes ratio
96  double bin_size = 0.0;
97  if (is_radial_)
98  bin_size = search_radius_ / image_width_;
99  else
100  bin_size = search_radius_ / image_width_ / sqrt(2.0);
101 
102  std::vector<int> nn_indices;
103  std::vector<float> nn_sqr_dists;
104  const int neighb_cnt = this->searchForNeighbors (index, search_radius_, nn_indices, nn_sqr_dists);
105  if (neighb_cnt < static_cast<int> (min_pts_neighb_))
106  {
107  throw PCLException (
108  "Too few points for spin image, use setMinPointCountInNeighbourhood() to decrease the threshold or use larger feature radius",
109  "spin_image.hpp", "computeSiForPoint");
110  }
111 
112  // for all neighbor points
113  for (int i_neigh = 0; i_neigh < neighb_cnt ; i_neigh++)
114  {
115  // first, skip the points with distant normals
116  double cos_between_normals = -2.0; // should be initialized if used
117  if (support_angle_cos_ > 0.0 || is_angular_) // not bogus
118  {
119  cos_between_normals = origin_normal.dot (input_normals_->points[nn_indices[i_neigh]].getNormalVector3fMap ());
120  if (fabs (cos_between_normals) > (1.0 + 10*std::numeric_limits<float>::epsilon ())) // should be okay for numeric stability
121  {
122  PCL_ERROR ("[pcl::%s::computeSiForPoint] Normal for the point %d and/or the point %d are not normalized, dot ptoduct is %f.\n",
123  getClassName ().c_str (), nn_indices[i_neigh], index, cos_between_normals);
124  throw PCLException ("Some normals are not normalized",
125  "spin_image.hpp", "computeSiForPoint");
126  }
127  cos_between_normals = std::max (-1.0, std::min (1.0, cos_between_normals));
128 
129  if (fabs (cos_between_normals) < support_angle_cos_ ) // allow counter-directed normals
130  {
131  continue;
132  }
133 
134  if (cos_between_normals < 0.0)
135  {
136  cos_between_normals = -cos_between_normals; // the normal is not used explicitly from now
137  }
138  }
139 
140  // now compute the coordinate in cylindric coordinate system associated with the origin point
141  const Eigen::Vector3f direction (
142  surface_->points[nn_indices[i_neigh]].getVector3fMap () - origin_point);
143  const double direction_norm = direction.norm ();
144  if (fabs(direction_norm) < 10*std::numeric_limits<double>::epsilon ())
145  continue; // ignore the point itself; it does not contribute really
146  assert (direction_norm > 0.0);
147 
148  // the angle between the normal vector and the direction to the point
149  double cos_dir_axis = direction.dot(rotation_axis) / direction_norm;
150  if (fabs(cos_dir_axis) > (1.0 + 10*std::numeric_limits<float>::epsilon())) // should be okay for numeric stability
151  {
152  PCL_ERROR ("[pcl::%s::computeSiForPoint] Rotation axis for the point %d are not normalized, dot ptoduct is %f.\n",
153  getClassName ().c_str (), index, cos_dir_axis);
154  throw PCLException ("Some rotation axis is not normalized",
155  "spin_image.hpp", "computeSiForPoint");
156  }
157  cos_dir_axis = std::max (-1.0, std::min (1.0, cos_dir_axis));
158 
159  // compute coordinates w.r.t. the reference frame
160  double beta = std::numeric_limits<double>::signaling_NaN ();
161  double alpha = std::numeric_limits<double>::signaling_NaN ();
162  if (is_radial_) // radial spin image structure
163  {
164  beta = asin (cos_dir_axis); // yes, arc sine! to get the angle against tangent, not normal!
165  alpha = direction_norm;
166  }
167  else // rectangular spin-image structure
168  {
169  beta = direction_norm * cos_dir_axis;
170  alpha = direction_norm * sqrt (1.0 - cos_dir_axis*cos_dir_axis);
171 
172  if (fabs (beta) >= bin_size * image_width_ || alpha >= bin_size * image_width_)
173  {
174  continue; // outside the cylinder
175  }
176  }
177 
178  assert (alpha >= 0.0);
179  assert (alpha <= bin_size * image_width_ + 20 * std::numeric_limits<float>::epsilon () );
180 
181 
182  // bilinear interpolation
183  double beta_bin_size = is_radial_ ? (M_PI / 2 / image_width_) : bin_size;
184  int beta_bin = int(std::floor (beta / beta_bin_size)) + int(image_width_);
185  assert (0 <= beta_bin && beta_bin < m_matrix.cols ());
186  int alpha_bin = int(std::floor (alpha / bin_size));
187  assert (0 <= alpha_bin && alpha_bin < m_matrix.rows ());
188 
189  if (alpha_bin == static_cast<int> (image_width_)) // border points
190  {
191  alpha_bin--;
192  // HACK: to prevent a > 1
193  alpha = bin_size * (alpha_bin + 1) - std::numeric_limits<double>::epsilon ();
194  }
195  if (beta_bin == int(2*image_width_) ) // border points
196  {
197  beta_bin--;
198  // HACK: to prevent b > 1
199  beta = beta_bin_size * (beta_bin - int(image_width_) + 1) - std::numeric_limits<double>::epsilon ();
200  }
201 
202  double a = alpha/bin_size - double(alpha_bin);
203  double b = beta/beta_bin_size - double(beta_bin-int(image_width_));
204 
205  assert (0 <= a && a <= 1);
206  assert (0 <= b && b <= 1);
207 
208  m_matrix (alpha_bin, beta_bin) += (1-a) * (1-b);
209  m_matrix (alpha_bin+1, beta_bin) += a * (1-b);
210  m_matrix (alpha_bin, beta_bin+1) += (1-a) * b;
211  m_matrix (alpha_bin+1, beta_bin+1) += a * b;
212 
213  if (is_angular_)
214  {
215  m_averAngles (alpha_bin, beta_bin) += (1-a) * (1-b) * acos (cos_between_normals);
216  m_averAngles (alpha_bin+1, beta_bin) += a * (1-b) * acos (cos_between_normals);
217  m_averAngles (alpha_bin, beta_bin+1) += (1-a) * b * acos (cos_between_normals);
218  m_averAngles (alpha_bin+1, beta_bin+1) += a * b * acos (cos_between_normals);
219  }
220  }
221 
222  if (is_angular_)
223  {
224  // transform sum to average
225  m_matrix = m_averAngles / (m_matrix + std::numeric_limits<double>::epsilon ()); // +eps to avoid division by zero
226  }
227  else if (neighb_cnt > 1) // to avoid division by zero, also no need to divide by 1
228  {
229  // normalization
230  m_matrix /= m_matrix.sum();
231  }
232 
233  return m_matrix;
234 }
235 
236 
237 //////////////////////////////////////////////////////////////////////////////////////////////
238 template <typename PointInT, typename PointNT, typename PointOutT> bool
240 {
242  {
243  PCL_ERROR ("[pcl::%s::initCompute] Init failed.\n", getClassName ().c_str ());
244  return (false);
245  }
246 
247  // Check if input normals are set
248  if (!input_normals_)
249  {
250  PCL_ERROR ("[pcl::%s::initCompute] No input dataset containing normals was given!\n", getClassName ().c_str ());
252  return (false);
253  }
254 
255  // Check if the size of normals is the same as the size of the surface
256  if (input_normals_->points.size () != input_->points.size ())
257  {
258  PCL_ERROR ("[pcl::%s::initCompute] ", getClassName ().c_str ());
259  PCL_ERROR ("The number of points in the input dataset differs from ");
260  PCL_ERROR ("the number of points in the dataset containing the normals!\n");
262  return (false);
263  }
264 
265  // We need a positive definite search radius to continue
266  if (search_radius_ == 0)
267  {
268  PCL_ERROR ("[pcl::%s::initCompute] Need a search radius different than 0!\n", getClassName ().c_str ());
270  return (false);
271  }
272  if (k_ != 0)
273  {
274  PCL_ERROR ("[pcl::%s::initCompute] K-nearest neighbor search for spin images not implemented. Used a search radius instead!\n", getClassName ().c_str ());
276  return (false);
277  }
278  // If the surface won't be set, make fake surface and fake surface normals
279  // if we wouldn't do it here, the following method would alarm that no surface normals is given
280  if (!surface_)
281  {
282  surface_ = input_;
283  fake_surface_ = true;
284  }
285 
286  //if (fake_surface_ && !input_normals_)
287  // input_normals_ = normals_; // normals_ is set, as checked earlier
288 
289  assert(!(use_custom_axis_ && use_custom_axes_cloud_));
290 
291  if (!use_custom_axis_ && !use_custom_axes_cloud_ // use input normals as rotation axes
292  && !input_normals_)
293  {
294  PCL_ERROR ("[pcl::%s::initCompute] No normals for input cloud were given!\n", getClassName ().c_str ());
295  // Cleanup
297  return (false);
298  }
299 
300  if ((is_angular_ || support_angle_cos_ > 0.0) // support angle is not bogus NOTE this is for randomly-flipped normals
301  && !input_normals_)
302  {
303  PCL_ERROR ("[pcl::%s::initCompute] No normals for input cloud were given!\n", getClassName ().c_str ());
304  // Cleanup
306  return (false);
307  }
308 
309  if (use_custom_axes_cloud_
310  && rotation_axes_cloud_->size () == input_->size ())
311  {
312  PCL_ERROR ("[pcl::%s::initCompute] Rotation axis cloud have different size from input!\n", getClassName ().c_str ());
313  // Cleanup
315  return (false);
316  }
317 
318  return (true);
319 }
320 
321 
322 //////////////////////////////////////////////////////////////////////////////////////////////
323 template <typename PointInT, typename PointNT, typename PointOutT> void
325 {
326  for (int i_input = 0; i_input < static_cast<int> (indices_->size ()); ++i_input)
327  {
328  Eigen::ArrayXXd res = computeSiForPoint (indices_->at (i_input));
329 
330  // Copy into the resultant cloud
331  for (int iRow = 0; iRow < res.rows () ; iRow++)
332  {
333  for (int iCol = 0; iCol < res.cols () ; iCol++)
334  {
335  output.points[i_input].histogram[ iRow*res.cols () + iCol ] = static_cast<float> (res (iRow, iCol));
336  }
337  }
338  }
339 }
340 
341 #define PCL_INSTANTIATE_SpinImageEstimation(T,NT,OutT) template class PCL_EXPORTS pcl::SpinImageEstimation<T,NT,OutT>;
342 
343 #endif // PCL_FEATURES_IMPL_SPIN_IMAGE_H_
344