How to extract NARF Features from a range image

This tutorial demonstrates how to extract NARF descriptors at NARF keypoint positions from a range image. The executable enables us to load a point cloud from disc (or create it if not given), extract interest points on it and then calculate the descriptors at these positions. It then visualizes these positions, both in an image and a 3D viewer.

The code

First, create a file called, let’s say, narf_feature_extraction.cpp in your favorite editor, and place the following code inside 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
 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
/* \author Bastian Steder */

#include <iostream>

#include <boost/thread/thread.hpp>
#include <pcl/range_image/range_image.h>
#include <pcl/io/pcd_io.h>
#include <pcl/visualization/range_image_visualizer.h>
#include <pcl/visualization/pcl_visualizer.h>
#include <pcl/features/range_image_border_extractor.h>
#include <pcl/keypoints/narf_keypoint.h>
#include <pcl/features/narf_descriptor.h>
#include <pcl/console/parse.h>

typedef pcl::PointXYZ PointType;

// --------------------
// -----Parameters-----
// --------------------
float angular_resolution = 0.5f;
float support_size = 0.2f;
pcl::RangeImage::CoordinateFrame coordinate_frame = pcl::RangeImage::CAMERA_FRAME;
bool setUnseenToMaxRange = false;
bool rotation_invariant = true;

// --------------
// -----Help-----
// --------------
void 
printUsage (const char* progName)
{
  std::cout << "\n\nUsage: "<<progName<<" [options] <scene.pcd>\n\n"
            << "Options:\n"
            << "-------------------------------------------\n"
            << "-r <float>   angular resolution in degrees (default "<<angular_resolution<<")\n"
            << "-c <int>     coordinate frame (default "<< (int)coordinate_frame<<")\n"
            << "-m           Treat all unseen points to max range\n"
            << "-s <float>   support size for the interest points (diameter of the used sphere - "
                                                                  "default "<<support_size<<")\n"
            << "-o <0/1>     switch rotational invariant version of the feature on/off"
            <<               " (default "<< (int)rotation_invariant<<")\n"
            << "-h           this help\n"
            << "\n\n";
}

void 
setViewerPose (pcl::visualization::PCLVisualizer& viewer, const Eigen::Affine3f& viewer_pose)
{
  Eigen::Vector3f pos_vector = viewer_pose * Eigen::Vector3f (0, 0, 0);
  Eigen::Vector3f look_at_vector = viewer_pose.rotation () * Eigen::Vector3f (0, 0, 1) + pos_vector;
  Eigen::Vector3f up_vector = viewer_pose.rotation () * Eigen::Vector3f (0, -1, 0);
  viewer.setCameraPosition (pos_vector[0], pos_vector[1], pos_vector[2],
                            look_at_vector[0], look_at_vector[1], look_at_vector[2],
                            up_vector[0], up_vector[1], up_vector[2]);
}

// --------------
// -----Main-----
// --------------
int 
main (int argc, char** argv)
{
  // --------------------------------------
  // -----Parse Command Line Arguments-----
  // --------------------------------------
  if (pcl::console::find_argument (argc, argv, "-h") >= 0)
  {
    printUsage (argv[0]);
    return 0;
  }
  if (pcl::console::find_argument (argc, argv, "-m") >= 0)
  {
    setUnseenToMaxRange = true;
    cout << "Setting unseen values in range image to maximum range readings.\n";
  }
  if (pcl::console::parse (argc, argv, "-o", rotation_invariant) >= 0)
    cout << "Switching rotation invariant feature version "<< (rotation_invariant ? "on" : "off")<<".\n";
  int tmp_coordinate_frame;
  if (pcl::console::parse (argc, argv, "-c", tmp_coordinate_frame) >= 0)
  {
    coordinate_frame = pcl::RangeImage::CoordinateFrame (tmp_coordinate_frame);
    cout << "Using coordinate frame "<< (int)coordinate_frame<<".\n";
  }
  if (pcl::console::parse (argc, argv, "-s", support_size) >= 0)
    cout << "Setting support size to "<<support_size<<".\n";
  if (pcl::console::parse (argc, argv, "-r", angular_resolution) >= 0)
    cout << "Setting angular resolution to "<<angular_resolution<<"deg.\n";
  angular_resolution = pcl::deg2rad (angular_resolution);
  
  // ------------------------------------------------------------------
  // -----Read pcd file or create example point cloud if not given-----
  // ------------------------------------------------------------------
  pcl::PointCloud<PointType>::Ptr point_cloud_ptr (new pcl::PointCloud<PointType>);
  pcl::PointCloud<PointType>& point_cloud = *point_cloud_ptr;
  pcl::PointCloud<pcl::PointWithViewpoint> far_ranges;
  Eigen::Affine3f scene_sensor_pose (Eigen::Affine3f::Identity ());
  std::vector<int> pcd_filename_indices = pcl::console::parse_file_extension_argument (argc, argv, "pcd");
  if (!pcd_filename_indices.empty ())
  {
    std::string filename = argv[pcd_filename_indices[0]];
    if (pcl::io::loadPCDFile (filename, point_cloud) == -1)
    {
      cerr << "Was not able to open file \""<<filename<<"\".\n";
      printUsage (argv[0]);
      return 0;
    }
    scene_sensor_pose = Eigen::Affine3f (Eigen::Translation3f (point_cloud.sensor_origin_[0],
                                                               point_cloud.sensor_origin_[1],
                                                               point_cloud.sensor_origin_[2])) *
                        Eigen::Affine3f (point_cloud.sensor_orientation_);
    std::string far_ranges_filename = pcl::getFilenameWithoutExtension (filename)+"_far_ranges.pcd";
    if (pcl::io::loadPCDFile (far_ranges_filename.c_str (), far_ranges) == -1)
      std::cout << "Far ranges file \""<<far_ranges_filename<<"\" does not exists.\n";
  }
  else
  {
    setUnseenToMaxRange = true;
    cout << "\nNo *.pcd file given => Genarating example point cloud.\n\n";
    for (float x=-0.5f; x<=0.5f; x+=0.01f)
    {
      for (float y=-0.5f; y<=0.5f; y+=0.01f)
      {
        PointType point;  point.x = x;  point.y = y;  point.z = 2.0f - y;
        point_cloud.points.push_back (point);
      }
    }
    point_cloud.width = (int) point_cloud.points.size ();  point_cloud.height = 1;
  }
  
  // -----------------------------------------------
  // -----Create RangeImage from the PointCloud-----
  // -----------------------------------------------
  float noise_level = 0.0;
  float min_range = 0.0f;
  int border_size = 1;
  boost::shared_ptr<pcl::RangeImage> range_image_ptr (new pcl::RangeImage);
  pcl::RangeImage& range_image = *range_image_ptr;   
  range_image.createFromPointCloud (point_cloud, angular_resolution, pcl::deg2rad (360.0f), pcl::deg2rad (180.0f),
                                   scene_sensor_pose, coordinate_frame, noise_level, min_range, border_size);
  range_image.integrateFarRanges (far_ranges);
  if (setUnseenToMaxRange)
    range_image.setUnseenToMaxRange ();
  
  // --------------------------------------------
  // -----Open 3D viewer and add point cloud-----
  // --------------------------------------------
  pcl::visualization::PCLVisualizer viewer ("3D Viewer");
  viewer.setBackgroundColor (1, 1, 1);
  pcl::visualization::PointCloudColorHandlerCustom<pcl::PointWithRange> range_image_color_handler (range_image_ptr, 0, 0, 0);
  viewer.addPointCloud (range_image_ptr, range_image_color_handler, "range image");
  viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 1, "range image");
  //viewer.addCoordinateSystem (1.0f, "global");
  //PointCloudColorHandlerCustom<PointType> point_cloud_color_handler (point_cloud_ptr, 150, 150, 150);
  //viewer.addPointCloud (point_cloud_ptr, point_cloud_color_handler, "original point cloud");
  viewer.initCameraParameters ();
  setViewerPose (viewer, range_image.getTransformationToWorldSystem ());
  
  // --------------------------
  // -----Show range image-----
  // --------------------------
  pcl::visualization::RangeImageVisualizer range_image_widget ("Range image");
  range_image_widget.showRangeImage (range_image);
  
  // --------------------------------
  // -----Extract NARF keypoints-----
  // --------------------------------
  pcl::RangeImageBorderExtractor range_image_border_extractor;
  pcl::NarfKeypoint narf_keypoint_detector;
  narf_keypoint_detector.setRangeImageBorderExtractor (&range_image_border_extractor);
  narf_keypoint_detector.setRangeImage (&range_image);
  narf_keypoint_detector.getParameters ().support_size = support_size;
  
  pcl::PointCloud<int> keypoint_indices;
  narf_keypoint_detector.compute (keypoint_indices);
  std::cout << "Found "<<keypoint_indices.points.size ()<<" key points.\n";

  // ----------------------------------------------
  // -----Show keypoints in range image widget-----
  // ----------------------------------------------
  //for (size_t i=0; i<keypoint_indices.points.size (); ++i)
    //range_image_widget.markPoint (keypoint_indices.points[i]%range_image.width,
                                  //keypoint_indices.points[i]/range_image.width);
  
  // -------------------------------------
  // -----Show keypoints in 3D viewer-----
  // -------------------------------------
  pcl::PointCloud<pcl::PointXYZ>::Ptr keypoints_ptr (new pcl::PointCloud<pcl::PointXYZ>);
  pcl::PointCloud<pcl::PointXYZ>& keypoints = *keypoints_ptr;
  keypoints.points.resize (keypoint_indices.points.size ());
  for (size_t i=0; i<keypoint_indices.points.size (); ++i)
    keypoints.points[i].getVector3fMap () = range_image.points[keypoint_indices.points[i]].getVector3fMap ();
  pcl::visualization::PointCloudColorHandlerCustom<pcl::PointXYZ> keypoints_color_handler (keypoints_ptr, 0, 255, 0);
  viewer.addPointCloud<pcl::PointXYZ> (keypoints_ptr, keypoints_color_handler, "keypoints");
  viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 7, "keypoints");
  
  // ------------------------------------------------------
  // -----Extract NARF descriptors for interest points-----
  // ------------------------------------------------------
  std::vector<int> keypoint_indices2;
  keypoint_indices2.resize (keypoint_indices.points.size ());
  for (unsigned int i=0; i<keypoint_indices.size (); ++i) // This step is necessary to get the right vector type
    keypoint_indices2[i]=keypoint_indices.points[i];
  pcl::NarfDescriptor narf_descriptor (&range_image, &keypoint_indices2);
  narf_descriptor.getParameters ().support_size = support_size;
  narf_descriptor.getParameters ().rotation_invariant = rotation_invariant;
  pcl::PointCloud<pcl::Narf36> narf_descriptors;
  narf_descriptor.compute (narf_descriptors);
  cout << "Extracted "<<narf_descriptors.size ()<<" descriptors for "
                      <<keypoint_indices.points.size ()<< " keypoints.\n";
  
  //--------------------
  // -----Main loop-----
  //--------------------
  while (!viewer.wasStopped ())
  {
    range_image_widget.spinOnce ();  // process GUI events
    viewer.spinOnce ();
    pcl_sleep(0.01);
  }
}

Explanation

In the beginning we do command line parsing, read a point cloud from disc (or create it if not provided), create a range image and extract NARF keypoints from it. All of these steps are already covered in the previous tutorial NARF keypoint extraction.

The interesting part begins here:

...
std::vector<int> keypoint_indices2;
keypoint_indices2.resize(keypoint_indices.points.size());
for (unsigned int i=0; i<keypoint_indices.size(); ++i) // This step is necessary to get the right vector type
  keypoint_indices2[i]=keypoint_indices.points[i];
...

Here we copy the indices to the vector used as input for the feature.

...
pcl::NarfDescriptor narf_descriptor(&range_image, &keypoint_indices2);
narf_descriptor.getParameters().support_size = support_size;
narf_descriptor.getParameters().rotation_invariant = rotation_invariant;
pcl::PointCloud<pcl::Narf36> narf_descriptors;
narf_descriptor.compute(narf_descriptors);
cout << "Extracted "<<narf_descriptors.size()<<" descriptors for "<<keypoint_indices.points.size()<< " keypoints.\n";
...

This code does the actual calculation of the descriptors. It first creates the NarfDescriptor object and gives it the input data (the keypoint indices and the range image). Then two important parameters are set. The support size, which determines the size of the area from which the descriptor is calculated, and if the rotational invariant (rotation around the normal) version of the NARF descriptor should be used. The we create the output pointcloud and do the actual computation. At last, we output the number of keypoints and the number of extracted descriptors. This numbers can differ. For one, it might happen that the calculation of the descriptor fails, because there are not enough points in the range image (resolution too low). Or there might be multiple descriptors in the same place, but for different dominant rotations.

The resulting PointCloud contains the type Narf36 (see common/include/pcl/point_types.h) and store the descriptor as a 36 elements float and x,y,z,roll,pitch,yaw to describe the local coordinate frame at which the feature was extracted. The descriptors can now be compared, e.g., whith the Manhatten distance (sum of absolute differences).

The remaining code just visualizes the keypoint positions in a range image widget and also in a 3D viewer.

Compiling and running the program

Add the following lines to your CMakeLists.txt file:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
cmake_minimum_required(VERSION 2.6 FATAL_ERROR)

project(narf_feature_extraction)

find_package(PCL 1.3 REQUIRED)

include_directories(${PCL_INCLUDE_DIRS})
link_directories(${PCL_LIBRARY_DIRS})
add_definitions(${PCL_DEFINITIONS})

add_executable (narf_feature_extraction narf_feature_extraction.cpp)
target_link_libraries (narf_feature_extraction ${PCL_LIBRARIES})

After you have made the executable, you can run it. Simply do:

$ ./narf_feature_extraction -m

This will use an autogenerated point cloud of a rectangle floating in space. The key points are detected in the corners. The parameter -m is necessary, since the area around the rectangle is unseen and therefore the system can not detect it as a border. The option -m changes the unseen area to maximum range readings, thereby enabling the system to use these borders.

You can also try it with a point cloud file from your hard drive:

$ ./narf_feature_extraction <point_cloud.pcd>

The output should look similar to this:

_images/narf_keypoint_extraction.png