Cluster Recognition and 6DOF Pose Estimation using VFH descriptors

As previously described in Estimating VFH signatures for a set of points, Viewpoint Feature Histograms (VFH) are powerful meta-local descriptors, created for the purpose of recognition and pose estimation for clusters of points. We here refer to a cluster as a collection of 3D points, most of the time representing a particular object or part of a scene, obtained through some segmentation or detection mechanisms (please see Euclidean Cluster Extraction for an example).

Our goal here is not to provide an ultimate recognition tool, but rather a mechanism for obtaining candidates that could potentially be the cluster/object that is searched for, together with its 6DOF pose in space. With this in mind, we will be formulating the recognition problem as a nearest neighbor estimation problem. So given a set of training data, we will use efficient nearest neighbor search structures such as kd-trees and return a set of potential candidates with sorted distances to the query object, rather than an absolute “this is the object that we were searching for” kind of response. The reader can imagine that such a system becomes much more useful as we can explicitly reason about failures (false positives, or true negatives).

For the purpose of this tutorial, the application example could be formulated as follows:

  • Training stage:
    • given a scene with 1 object that is easily separable as a cluster;
    • use a ground-truth system to obtain its pose (see the discussion below);
    • rotate around the object or rotate the object with respect to the camera, and compute a VFH descriptor for each view;
    • store the views, and build a kd-tree representation.
  • Testing stage:
    • given a scene with objects that can be separated as individual clusters, first extract the clusters;
    • for each cluster, compute a VFH descriptor from the current camera position;
    • use the VFH descriptor to search for candidates in the trained kd-tree.

We hope the above makes sense. Basically we’re first going to create the set of objects that we try to later on recognize, and then we will use that to obtain valid candidates for objects in the scene.

A good example of a ground-truth system could be a simple rotating pan-tilt unit such as the one in the figure below. Placing an object on the unit, and moving it with some increments in both horizontal and vertical, can result in a perfect ground-truth system for small objects. A cheaper solution could be to use a marker-based system (e.g., checkerboard) and rotate the camera/table manually.

_images/pan_tilt.jpg

Our Kd-Tree implementation of choice for the purpose of this tutorial is of course, FLANN.

Training

We begin the training by assuming that the objects are already separated as individual clusters (see Euclidean Cluster Extraction), as shown in the figure below:

_images/scene_raw.jpg _images/scene_segmented.jpg

Since we’re only trying to cover the explicity training/testing of VFH signatures in this tutorial, we provide a set of datasets already collected at: vfh_recognition_tutorial_data.tbz. The data is a subset of the objects presented in the figure below (left), and look like the point clouds on the right. We used the pan-tilt table shown above to acquire the data.

_images/objects.jpg _images/training.jpg

Next, copy and paste the following code into your editor and save it as build_tree.cpp.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
#include <pcl/point_types.h>
#include <pcl/point_cloud.h>
#include <pcl/console/parse.h>
#include <pcl/console/print.h>
#include <pcl/io/pcd_io.h>
#include <boost/filesystem.hpp>
#include <flann/flann.h>
#include <flann/io/hdf5.h>
#include <fstream>

typedef std::pair<std::string, std::vector<float> > vfh_model;

/** \brief Loads an n-D histogram file as a VFH signature
  * \param path the input file name
  * \param vfh the resultant VFH model
  */
bool
loadHist (const boost::filesystem::path &path, vfh_model &vfh)
{
  int vfh_idx;
  // Load the file as a PCD
  try
  {
    pcl::PCLPointCloud2 cloud;
    int version;
    Eigen::Vector4f origin;
    Eigen::Quaternionf orientation;
    pcl::PCDReader r;
    int type; unsigned int idx;
    r.readHeader (path.string (), cloud, origin, orientation, version, type, idx);

    vfh_idx = pcl::getFieldIndex (cloud, "vfh");
    if (vfh_idx == -1)
      return (false);
    if ((int)cloud.width * cloud.height != 1)
      return (false);
  }
  catch (const pcl::InvalidConversionException&)
  {
    return (false);
  }

  // Treat the VFH signature as a single Point Cloud
  pcl::PointCloud <pcl::VFHSignature308> point;
  pcl::io::loadPCDFile (path.string (), point);
  vfh.second.resize (308);

  std::vector <pcl::PCLPointField> fields;
  pcl::getFieldIndex (point, "vfh", fields);

  for (size_t i = 0; i < fields[vfh_idx].count; ++i)
  {
    vfh.second[i] = point.points[0].histogram[i];
  }
  vfh.first = path.string ();
  return (true);
}

/** \brief Load a set of VFH features that will act as the model (training data)
  * \param argc the number of arguments (pass from main ())
  * \param argv the actual command line arguments (pass from main ())
  * \param extension the file extension containing the VFH features
  * \param models the resultant vector of histogram models
  */
void
loadFeatureModels (const boost::filesystem::path &base_dir, const std::string &extension, 
                   std::vector<vfh_model> &models)
{
  if (!boost::filesystem::exists (base_dir) && !boost::filesystem::is_directory (base_dir))
    return;

  for (boost::filesystem::directory_iterator it (base_dir); it != boost::filesystem::directory_iterator (); ++it)
  {
    if (boost::filesystem::is_directory (it->status ()))
    {
      std::stringstream ss;
      ss << it->path ();
      pcl::console::print_highlight ("Loading %s (%lu models loaded so far).\n", ss.str ().c_str (), (unsigned long)models.size ());
      loadFeatureModels (it->path (), extension, models);
    }
    if (boost::filesystem::is_regular_file (it->status ()) && boost::filesystem::extension (it->path ()) == extension)
    {
      vfh_model m;
      if (loadHist (base_dir / it->path ().filename (), m))
        models.push_back (m);
    }
  }
}

int
main (int argc, char** argv)
{
  if (argc < 2)
  {
    PCL_ERROR ("Need at least two parameters! Syntax is: %s [model_directory] [options]\n", argv[0]);
    return (-1);
  }

  std::string extension (".pcd");
  transform (extension.begin (), extension.end (), extension.begin (), (int(*)(int))tolower);

  std::string kdtree_idx_file_name = "kdtree.idx";
  std::string training_data_h5_file_name = "training_data.h5";
  std::string training_data_list_file_name = "training_data.list";

  std::vector<vfh_model> models;

  // Load the model histograms
  loadFeatureModels (argv[1], extension, models);
  pcl::console::print_highlight ("Loaded %d VFH models. Creating training data %s/%s.\n", 
      (int)models.size (), training_data_h5_file_name.c_str (), training_data_list_file_name.c_str ());

  // Convert data into FLANN format
  flann::Matrix<float> data (new float[models.size () * models[0].second.size ()], models.size (), models[0].second.size ());

  for (size_t i = 0; i < data.rows; ++i)
    for (size_t j = 0; j < data.cols; ++j)
      data[i][j] = models[i].second[j];

  // Save data to disk (list of models)
  flann::save_to_file (data, training_data_h5_file_name, "training_data");
  std::ofstream fs;
  fs.open (training_data_list_file_name.c_str ());
  for (size_t i = 0; i < models.size (); ++i)
    fs << models[i].first << "\n";
  fs.close ();
 
  // Build the tree index and save it to disk
  pcl::console::print_error ("Building the kdtree index (%s) for %d elements...\n", kdtree_idx_file_name.c_str (), (int)data.rows);
  flann::Index<flann::ChiSquareDistance<float> > index (data, flann::LinearIndexParams ());
  //flann::Index<flann::ChiSquareDistance<float> > index (data, flann::KDTreeIndexParams (4));
  index.buildIndex ();
  index.save (kdtree_idx_file_name);
  delete[] data.ptr ();

  return (0);
}

In the following paragraphs we will explain what the above code does (or should do). We’ll begin with the main function.

We begin by loading a set of feature models from a directory given as the first command line argument (see details for running the example below). The loadFeatureModels method does nothing but recursively traverse a set of directories and subdirectories, and loads in all .PCD files it finds. In loadFeatureModels, we call loadHist, which will attempt to open each PCD file found, read its header, and check whether it contains a VFH signature or not. Together with the VFH signature we also store the PCD file name into a vfh_model pair.

Once all VFH features have been loaded, we convert them to FLANN format, using:

  // Convert data into FLANN format
  flann::Matrix<float> data (new float[models.size () * models[0].second.size ()], models.size (), models[0].second.size ());

  for (size_t i = 0; i < data.rows; ++i)
    for (size_t j = 0; j < data.cols; ++j)
      data[i][j] = models[i].second[j];

Since we’re lazy, and we want to use this data (and not reload it again by crawling the directory structure in the testing phase), we dump the data to disk:

  // Save data to disk (list of models)
  flann::save_to_file (data, training_data_h5_file_name, "training_data");
  std::ofstream fs;
  fs.open (training_data_list_file_name.c_str ());
  for (size_t i = 0; i < models.size (); ++i)
    fs << models[i].first << "\n";
  fs.close ();

Finally, we create the KdTree, and save its structure to disk:

  pcl::console::print_error ("Building the kdtree index (%s) for %d elements...\n", kdtree_idx_file_name.c_str (), (int)data.rows);
  flann::Index<flann::ChiSquareDistance<float> > index (data, flann::LinearIndexParams ());
  //flann::Index<flann::ChiSquareDistance<float> > index (data, flann::KDTreeIndexParams (4));
  index.buildIndex ();
  index.save (kdtree_idx_file_name);

Here we will use a LinearIndex, which does a brute-force search using a Chi-Square distance metric (see [VFH] for more information). For building a proper kd-tree, comment line 1 and uncomment line 2 in the code snippet above. The most important difference between a LinearIndex and a KDTreeIndex in FLANN is that the KDTree will be much faster, while producing approximate nearest neighbor results, rather than absolute.

So, we’re done with training. To summarize:

  1. we crawled a directory structure, looked at all the .PCD files we found, tested them whether they are VFH signatures and loaded them in memory;
  2. we converted the data into FLANN format and dumped it to disk;
  3. we built a kd-tree structure and dumped it to disk.

Testing

In the testing phase, we will illustrate how the system works by randomly loading one of the files used in the training phase (feel free to supply your own file here!), and checking the results of the tree.

Begin by copying and pasting the following code into your editor and save it as nearest_neighbors.cpp.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
#include <pcl/point_types.h>
#include <pcl/point_cloud.h>
#include <pcl/common/common.h>
#include <pcl/common/transforms.h>
#include <pcl/visualization/pcl_visualizer.h>
#include <pcl/console/parse.h>
#include <pcl/console/print.h>
#include <pcl/io/pcd_io.h>
#include <iostream>
#include <flann/flann.h>
#include <flann/io/hdf5.h>
#include <boost/filesystem.hpp>

typedef std::pair<std::string, std::vector<float> > vfh_model;

/** \brief Loads an n-D histogram file as a VFH signature
  * \param path the input file name
  * \param vfh the resultant VFH model
  */
bool
loadHist (const boost::filesystem::path &path, vfh_model &vfh)
{
  int vfh_idx;
  // Load the file as a PCD
  try
  {
    pcl::PCLPointCloud2 cloud;
    int version;
    Eigen::Vector4f origin;
    Eigen::Quaternionf orientation;
    pcl::PCDReader r;
    int type; unsigned int idx;
    r.readHeader (path.string (), cloud, origin, orientation, version, type, idx);

    vfh_idx = pcl::getFieldIndex (cloud, "vfh");
    if (vfh_idx == -1)
      return (false);
    if ((int)cloud.width * cloud.height != 1)
      return (false);
  }
  catch (const pcl::InvalidConversionException&)
  {
    return (false);
  }

  // Treat the VFH signature as a single Point Cloud
  pcl::PointCloud <pcl::VFHSignature308> point;
  pcl::io::loadPCDFile (path.string (), point);
  vfh.second.resize (308);

  std::vector <pcl::PCLPointField> fields;
  getFieldIndex (point, "vfh", fields);

  for (size_t i = 0; i < fields[vfh_idx].count; ++i)
  {
    vfh.second[i] = point.points[0].histogram[i];
  }
  vfh.first = path.string ();
  return (true);
}


/** \brief Search for the closest k neighbors
  * \param index the tree
  * \param model the query model
  * \param k the number of neighbors to search for
  * \param indices the resultant neighbor indices
  * \param distances the resultant neighbor distances
  */
inline void
nearestKSearch (flann::Index<flann::ChiSquareDistance<float> > &index, const vfh_model &model, 
                int k, flann::Matrix<int> &indices, flann::Matrix<float> &distances)
{
  // Query point
  flann::Matrix<float> p = flann::Matrix<float>(new float[model.second.size ()], 1, model.second.size ());
  memcpy (&p.ptr ()[0], &model.second[0], p.cols * p.rows * sizeof (float));

  indices = flann::Matrix<int>(new int[k], 1, k);
  distances = flann::Matrix<float>(new float[k], 1, k);
  index.knnSearch (p, indices, distances, k, flann::SearchParams (512));
  delete[] p.ptr ();
}

/** \brief Load the list of file model names from an ASCII file
  * \param models the resultant list of model name
  * \param filename the input file name
  */
bool
loadFileList (std::vector<vfh_model> &models, const std::string &filename)
{
  ifstream fs;
  fs.open (filename.c_str ());
  if (!fs.is_open () || fs.fail ())
    return (false);

  std::string line;
  while (!fs.eof ())
  {
    getline (fs, line);
    if (line.empty ())
      continue;
    vfh_model m;
    m.first = line;
    models.push_back (m);
  }
  fs.close ();
  return (true);
}

int
main (int argc, char** argv)
{
  int k = 6;

  double thresh = DBL_MAX;     // No threshold, disabled by default

  if (argc < 2)
  {
    pcl::console::print_error 
      ("Need at least three parameters! Syntax is: %s <query_vfh_model.pcd> [options] {kdtree.idx} {training_data.h5} {training_data.list}\n", argv[0]);
    pcl::console::print_info ("    where [options] are:  -k      = number of nearest neighbors to search for in the tree (default: "); 
    pcl::console::print_value ("%d", k); pcl::console::print_info (")\n");
    pcl::console::print_info ("                          -thresh = maximum distance threshold for a model to be considered VALID (default: "); 
    pcl::console::print_value ("%f", thresh); pcl::console::print_info (")\n\n");
    return (-1);
  }

  std::string extension (".pcd");
  transform (extension.begin (), extension.end (), extension.begin (), (int(*)(int))tolower);

  // Load the test histogram
  std::vector<int> pcd_indices = pcl::console::parse_file_extension_argument (argc, argv, ".pcd");
  vfh_model histogram;
  if (!loadHist (argv[pcd_indices.at (0)], histogram))
  {
    pcl::console::print_error ("Cannot load test file %s\n", argv[pcd_indices.at (0)]);
    return (-1);
  }

  pcl::console::parse_argument (argc, argv, "-thresh", thresh);
  // Search for the k closest matches
  pcl::console::parse_argument (argc, argv, "-k", k);
  pcl::console::print_highlight ("Using "); pcl::console::print_value ("%d", k); pcl::console::print_info (" nearest neighbors.\n");

  std::string kdtree_idx_file_name = "kdtree.idx";
  std::string training_data_h5_file_name = "training_data.h5";
  std::string training_data_list_file_name = "training_data.list";

  std::vector<vfh_model> models;
  flann::Matrix<int> k_indices;
  flann::Matrix<float> k_distances;
  flann::Matrix<float> data;
  // Check if the data has already been saved to disk
  if (!boost::filesystem::exists ("training_data.h5") || !boost::filesystem::exists ("training_data.list"))
  {
    pcl::console::print_error ("Could not find training data models files %s and %s!\n", 
        training_data_h5_file_name.c_str (), training_data_list_file_name.c_str ());
    return (-1);
  }
  else
  {
    loadFileList (models, training_data_list_file_name);
    flann::load_from_file (data, training_data_h5_file_name, "training_data");
    pcl::console::print_highlight ("Training data found. Loaded %d VFH models from %s/%s.\n", 
        (int)data.rows, training_data_h5_file_name.c_str (), training_data_list_file_name.c_str ());
  }

  // Check if the tree index has already been saved to disk
  if (!boost::filesystem::exists (kdtree_idx_file_name))
  {
    pcl::console::print_error ("Could not find kd-tree index in file %s!", kdtree_idx_file_name.c_str ());
    return (-1);
  }
  else
  {
    flann::Index<flann::ChiSquareDistance<float> > index (data, flann::SavedIndexParams ("kdtree.idx"));
    index.buildIndex ();
    nearestKSearch (index, histogram, k, k_indices, k_distances);
  }

  // Output the results on screen
  pcl::console::print_highlight ("The closest %d neighbors for %s are:\n", k, argv[pcd_indices[0]]);
  for (int i = 0; i < k; ++i)
    pcl::console::print_info ("    %d - %s (%d) with a distance of: %f\n", 
        i, models.at (k_indices[0][i]).first.c_str (), k_indices[0][i], k_distances[0][i]);

  // Load the results
  pcl::visualization::PCLVisualizer p (argc, argv, "VFH Cluster Classifier");
  int y_s = (int)floor (sqrt ((double)k));
  int x_s = y_s + (int)ceil ((k / (double)y_s) - y_s);
  double x_step = (double)(1 / (double)x_s);
  double y_step = (double)(1 / (double)y_s);
  pcl::console::print_highlight ("Preparing to load "); 
  pcl::console::print_value ("%d", k); 
  pcl::console::print_info (" files ("); 
  pcl::console::print_value ("%d", x_s);    
  pcl::console::print_info ("x"); 
  pcl::console::print_value ("%d", y_s); 
  pcl::console::print_info (" / ");
  pcl::console::print_value ("%f", x_step); 
  pcl::console::print_info ("x"); 
  pcl::console::print_value ("%f", y_step); 
  pcl::console::print_info (")\n");

  int viewport = 0, l = 0, m = 0;
  for (int i = 0; i < k; ++i)
  {
    std::string cloud_name = models.at (k_indices[0][i]).first;
    boost::replace_last (cloud_name, "_vfh", "");

    p.createViewPort (l * x_step, m * y_step, (l + 1) * x_step, (m + 1) * y_step, viewport);
    l++;
    if (l >= x_s)
    {
      l = 0;
      m++;
    }

    pcl::PCLPointCloud2 cloud;
    pcl::console::print_highlight (stderr, "Loading "); pcl::console::print_value (stderr, "%s ", cloud_name.c_str ());
    if (pcl::io::loadPCDFile (cloud_name, cloud) == -1)
      break;

    // Convert from blob to PointCloud
    pcl::PointCloud<pcl::PointXYZ> cloud_xyz;
    pcl::fromPCLPointCloud2 (cloud, cloud_xyz);

    if (cloud_xyz.points.size () == 0)
      break;

    pcl::console::print_info ("[done, "); 
    pcl::console::print_value ("%d", (int)cloud_xyz.points.size ()); 
    pcl::console::print_info (" points]\n");
    pcl::console::print_info ("Available dimensions: "); 
    pcl::console::print_value ("%s\n", pcl::getFieldsList (cloud).c_str ());

    // Demean the cloud
    Eigen::Vector4f centroid;
    pcl::compute3DCentroid (cloud_xyz, centroid);
    pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_xyz_demean (new pcl::PointCloud<pcl::PointXYZ>);
    pcl::demeanPointCloud<pcl::PointXYZ> (cloud_xyz, centroid, *cloud_xyz_demean);
    // Add to renderer*
    p.addPointCloud (cloud_xyz_demean, cloud_name, viewport);
    
    // Check if the model found is within our inlier tolerance
    std::stringstream ss;
    ss << k_distances[0][i];
    if (k_distances[0][i] > thresh)
    {
      p.addText (ss.str (), 20, 30, 1, 0, 0, ss.str (), viewport);  // display the text with red

      // Create a red line
      pcl::PointXYZ min_p, max_p;
      pcl::getMinMax3D (*cloud_xyz_demean, min_p, max_p);
      std::stringstream line_name;
      line_name << "line_" << i;
      p.addLine (min_p, max_p, 1, 0, 0, line_name.str (), viewport);
      p.setShapeRenderingProperties (pcl::visualization::PCL_VISUALIZER_LINE_WIDTH, 5, line_name.str (), viewport);
    }
    else
      p.addText (ss.str (), 20, 30, 0, 1, 0, ss.str (), viewport);

    // Increase the font size for the score*
    p.setShapeRenderingProperties (pcl::visualization::PCL_VISUALIZER_FONT_SIZE, 18, ss.str (), viewport);

    // Add the cluster name
    p.addText (cloud_name, 20, 10, cloud_name, viewport);
  }
  // Add coordianate systems to all viewports
  p.addCoordinateSystem (0.1, "global", 0);

  p.spin ();
  return (0);
}

The above code snippet is slightly larger, because we also included some visualization routines and some other “eye candy” stuff.

In lines:

  std::vector<int> pcd_indices = pcl::console::parse_file_extension_argument (argc, argv, ".pcd");
  vfh_model histogram;
  if (!loadHist (argv[pcd_indices.at (0)], histogram))
  {
    pcl::console::print_error ("Cannot load test file %s\n", argv[pcd_indices.at (0)]);
    return (-1);
  }

  pcl::console::parse_argument (argc, argv, "-thresh", thresh);
  // Search for the k closest matches
  pcl::console::parse_argument (argc, argv, "-k", k);
  pcl::console::print_highlight ("Using "); pcl::console::print_value ("%d", k); pcl::console::print_info (" nearest neighbors.\n");

we load the first given user histogram (and ignore the rest). Then we proceed at checking two command line parameters, namely -k which will define how many nearest neighbors to check and display on screen, and -thresh which defines a maximum distance metric after which we will start displaying red lines (i.e., crossing) over the k models found on screen (eye candy!).

In lines:

    loadFileList (models, training_data_list_file_name);
    flann::load_from_file (data, training_data_h5_file_name, "training_data");

we load the training data from disk, together with the list of file names that we previously stored in build_tree.cpp. Then, we read the kd-tree and rebuild the index:

    flann::Index<flann::ChiSquareDistance<float> > index (data, flann::SavedIndexParams ("kdtree.idx"));
    index.buildIndex ();

Here we need to make sure that we use the exact distance metric (ChiSquareDistance in this case), as the one that we used while creating the tree. The most important part of the code comes here:

    nearestKSearch (index, histogram, k, k_indices, k_distances);

Inside nearestKSearch, we first convert the query point to FLANN format:

  flann::Matrix<float> p = flann::Matrix<float>(new float[model.second.size ()], 1, model.second.size ());
  memcpy (&p.ptr ()[0], &model.second[0], p.cols * p.rows * sizeof (float));

Followed by obtaining the resultant nearest neighbor indices and distances for the query in:

  indices = flann::Matrix<int>(new int[k], 1, k);
  distances = flann::Matrix<float>(new float[k], 1, k);
  index.knnSearch (p, indices, distances, k, flann::SearchParams (512));

Lines:

    index.buildIndex ();
    nearestKSearch (index, histogram, k, k_indices, k_distances);
  }

  // Output the results on screen
  pcl::console::print_highlight ("The closest %d neighbors for %s are:\n", k, argv[pcd_indices[0]]);
  for (int i = 0; i < k; ++i)
    pcl::console::print_info ("    %d - %s (%d) with a distance of: %f\n", 
        i, models.at (k_indices[0][i]).first.c_str (), k_indices[0][i], k_distances[0][i]);

  // Load the results
  pcl::visualization::PCLVisualizer p (argc, argv, "VFH Cluster Classifier");
  int y_s = (int)floor (sqrt ((double)k));
  int x_s = y_s + (int)ceil ((k / (double)y_s) - y_s);
  double x_step = (double)(1 / (double)x_s);

create a PCLVisualizer object, and sets up a set of different viewports (e.g., splits the screen into different chunks), which will be enabled in:

    p.createViewPort (l * x_step, m * y_step, (l + 1) * x_step, (m + 1) * y_step, viewport);

Using the file names representing the models that we previously obtained in loadFileList, we proceed at loading the model file names using:

    pcl::PCLPointCloud2 cloud;
    pcl::console::print_highlight (stderr, "Loading "); pcl::console::print_value (stderr, "%s ", cloud_name.c_str ());
    if (pcl::io::loadPCDFile (cloud_name, cloud) == -1)
      break;

    // Convert from blob to PointCloud
    pcl::PointCloud<pcl::PointXYZ> cloud_xyz;
    pcl::fromPCLPointCloud2 (cloud, cloud_xyz);

For visualization purposes, we demean the point cloud by computing its centroid and then subtracting it:

    Eigen::Vector4f centroid;
    pcl::compute3DCentroid (cloud_xyz, centroid);
    pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_xyz_demean (new pcl::PointCloud<pcl::PointXYZ>);
    pcl::demeanPointCloud<pcl::PointXYZ> (cloud_xyz, centroid, *cloud_xyz_demean);
    // Add to renderer*
    p.addPointCloud (cloud_xyz_demean, cloud_name, viewport);

Finally we check if the distance obtained by nearestKSearch is larger than the user given threshold, and if it is, we display a red line over the cloud that is being rendered in the viewport:

      // Create a red line
      pcl::PointXYZ min_p, max_p;
      pcl::getMinMax3D (*cloud_xyz_demean, min_p, max_p);
      std::stringstream line_name;
      line_name << "line_" << i;
      p.addLine (min_p, max_p, 1, 0, 0, line_name.str (), viewport);
      p.setShapeRenderingProperties (pcl::visualization::PCL_VISUALIZER_LINE_WIDTH, 5, line_name.str (), viewport);

Compiling and running the code

Create a new CMakeLists.txt file, and put the following content into it

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
cmake_minimum_required(VERSION 2.8 FATAL_ERROR)

# we need FindFLANN.cmake 
list(APPEND CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR})

project(vfh_cluster_classifier)

find_package(PCL 1.2 REQUIRED)
include_directories(${PCL_INCLUDE_DIRS})
link_directories(${PCL_LIBRARY_DIRS})
add_definitions(${PCL_DEFINITIONS})

find_package(HDF5)
if(HDF5_FOUND)

  find_package(FLANN)
  include_directories(${FLANN_INCLUDE_DIRS})

  include_directories(${HDF5_INCLUDE_DIR})

  add_executable(build_tree build_tree.cpp)
  target_link_libraries(build_tree ${PCL_LIBRARIES} ${Boost_LIBRARIES}
                                 ${FLANN_LIBRARIES} ${HDF5_hdf5_LIBRARY})

  add_executable(nearest_neighbors nearest_neighbors.cpp)
  target_link_libraries(nearest_neighbors ${PCL_LIBRARIES} 
                                        ${Boost_LIBRARIES} ${FLANN_LIBRARIES} ${HDF5_hdf5_LIBRARY} 
                                        )
endif(HDF5_FOUND)

Note

If you are running this tutorial on Windows, you have to install (HDF5 1.8.7 Shared Library). If CMake is not able to find HDF5, you can manually supply the include directory in HDF5_INCLUDE_DIR variable and the full path of hdf5dll.lib in HDF5_hdf5_LIBRARY variable. Make sure that the needed dlls are in the same folder as the executables.

The above assumes that your two source files (build_tree.cpp and nearest_neighbors.cpp) are stored into the src/ subdirectory.

Then, make sure that the datasets you downloaded (vfh_recognition_tutorial_data.tbz) are unpacked in this directory, thus creating a data/ subdirectory.

After you have made the executable, you can run them like so:

$ ./build/build_tree data/

You should see the following output on screen:

> Loading data/001.324.25 (0 models loaded so far).
> Loading data/800.919.49 (13 models loaded so far).
> Loading data/100.922.16 (27 models loaded so far).
> Loading data/901.125.07 (47 models loaded so far).
> Loading data/000.580.67 (65 models loaded so far).
> Loading data/463.156.00 (81 models loaded so far).
> Loading data/401.431.44 (97 models loaded so far).
> Loading data/100.919.00 (113 models loaded so far).
> Loading data/401.324.52 (134 models loaded so far).
> Loading data/201.327.78 (150 models loaded so far).
> Loading data/300.151.23 (166 models loaded so far).
> Loading data/200.921.07 (180 models loaded so far).
> Loaded 195 VFH models. Creating training data training_data.h5/training_data.list.
Building the kdtree index (kdtree.idx) for 195 elements...

The above crawled the data/ subdirectory, and created a kd-tree with 195 entries. To run the nearest neighbor testing example, you have two options:

  1. Either run the following command manually, and select one of the datasets that we provided as a testing sample, like this:

    ./build/nearest_neighbors -k 16 -thresh 50 data/000.580.67/1258730231333_cluster_0_nxyz_vfh.pcd
    
  2. Or, if you are on a linux system, you can place the following on a bash script file (e.g., test.sh):

    #!/bin/bash
    
    # Example directory containing _vfh.pcd files
    DATA=data
    
    # Inlier distance threshold
    thresh=50
    
    # Get the closest K nearest neighbors
    k=16
    
    for i in `find $DATA -type d -name "*"`
    do
      echo $i
      for j in `find $i -type f \( -iname "*cluster*_vfh.pcd" \) | sort -R`
      do
        echo $j
        ./build/nearest_neighbors -k $k -thresh $thresh $j -cam "0.403137,0.868471/0,0,0/-0.0932051,-0.201608,-0.518939/-0.00471487,-0.931831,0.362863/1464,764/6,72"
      done
    done
    

and run the script like this:

bash test.sh

You should see recognition examples like the ones shown below:

_images/vfh_example1.jpg _images/vfh_example2.jpg _images/vfh_example3.jpg