Visual Servoing Platform  version 3.5.0 under development (2022-02-15)
tutorial-apriltag-detector-live-rgbd-structure-core.cpp
1 #include <visp3/core/vpConfig.h>
3 #ifdef VISP_HAVE_MODULE_SENSOR
4 #include <visp3/sensor/vpOccipitalStructure.h>
5 #endif
6 #include <visp3/detection/vpDetectorAprilTag.h>
9 #include <visp3/gui/vpDisplayGDI.h>
10 #include <visp3/gui/vpDisplayOpenCV.h>
11 #include <visp3/gui/vpDisplayX.h>
12 #include <visp3/core/vpXmlParserCamera.h>
13 #include <visp3/core/vpImageConvert.h>
14 #include <visp3/vision/vpPose.h>
15 
16 int main(int argc, const char **argv)
17 {
19 #if defined(VISP_HAVE_APRILTAG) && defined(VISP_HAVE_OCCIPITAL_STRUCTURE)
20 
24  double tagSize = 0.053;
25  float quad_decimate = 1.0;
26  int nThreads = 1;
27  bool display_tag = false;
28  int color_id = -1;
29  unsigned int thickness = 2;
30  bool align_frame = false;
31 
32 #if !(defined(VISP_HAVE_X11) || defined(VISP_HAVE_GDI) || defined(VISP_HAVE_OPENCV))
33  bool display_off = true;
34  std::cout << "Warning: There is no 3rd party (X11, GDI or openCV) to dislay images..." << std::endl;
35 #else
36  bool display_off = false;
37 #endif
38 
39  for (int i = 1; i < argc; i++) {
40  if (std::string(argv[i]) == "--pose_method" && i + 1 < argc) {
41  poseEstimationMethod = (vpDetectorAprilTag::vpPoseEstimationMethod)atoi(argv[i + 1]);
42  } else if (std::string(argv[i]) == "--tag_size" && i + 1 < argc) {
43  tagSize = atof(argv[i + 1]);
44  } else if (std::string(argv[i]) == "--quad_decimate" && i + 1 < argc) {
45  quad_decimate = (float)atof(argv[i + 1]);
46  } else if (std::string(argv[i]) == "--nthreads" && i + 1 < argc) {
47  nThreads = atoi(argv[i + 1]);
48  } else if (std::string(argv[i]) == "--display_tag") {
49  display_tag = true;
50  } else if (std::string(argv[i]) == "--display_off") {
51  display_off = true;
52  } else if (std::string(argv[i]) == "--color" && i + 1 < argc) {
53  color_id = atoi(argv[i+1]);
54  } else if (std::string(argv[i]) == "--thickness" && i + 1 < argc) {
55  thickness = (unsigned int) atoi(argv[i+1]);
56  } else if (std::string(argv[i]) == "--tag_family" && i + 1 < argc) {
57  tagFamily = (vpDetectorAprilTag::vpAprilTagFamily)atoi(argv[i + 1]);
58  } else if (std::string(argv[i]) == "--z_aligned") {
59  align_frame = true;
60  }
61  else if (std::string(argv[i]) == "--help" || std::string(argv[i]) == "-h") {
62  std::cout << "Usage: " << argv[0]
63  << " [--tag_size <tag_size in m> (default: 0.053)]"
64  " [--quad_decimate <quad_decimate> (default: 1)]"
65  " [--nthreads <nb> (default: 1)]"
66  " [--pose_method <method> (0: HOMOGRAPHY, 1: HOMOGRAPHY_VIRTUAL_VS, "
67  " 2: DEMENTHON_VIRTUAL_VS, 3: LAGRANGE_VIRTUAL_VS, "
68  " 4: BEST_RESIDUAL_VIRTUAL_VS, 5: HOMOGRAPHY_ORTHOGONAL_ITERATION) (default: 0)]"
69  " [--tag_family <family> (0: TAG_36h11, 1: TAG_36h10 (DEPRECATED), 2: TAG_36ARTOOLKIT (DEPRECATED),"
70  " 3: TAG_25h9, 4: TAG_25h7 (DEPRECATED), 5: TAG_16h5, 6: TAG_CIRCLE21h7, 7: TAG_CIRCLE49h12,"
71  " 8: TAG_CUSTOM48h12, 9: TAG_STANDARD41h12, 10: TAG_STANDARD52h13) (default: 0)]"
72  " [--display_tag] [--z_aligned]";
73 #if (defined(VISP_HAVE_X11) || defined(VISP_HAVE_GDI) || defined(VISP_HAVE_OPENCV))
74  std::cout << " [--display_off] [--color <color id>] [--thickness <line thickness>]";
75 #endif
76  std::cout << " [--help]" << std::endl;
77  return EXIT_SUCCESS;
78  }
79  }
80 
81  try {
83  std::cout << "Use Occipital Structure grabber" << std::endl;
85  ST::CaptureSessionSettings settings;
86  settings.source = ST::CaptureSessionSourceId::StructureCore;
87  settings.structureCore.visibleEnabled = true;
88  settings.applyExpensiveCorrection = true; // Apply a correction and clean filter to the depth before streaming.
89  unsigned int width = 640, height = 480;
90 
92  vpImage<vpRGBa> I_color(height, width);
93  vpImage<float> I_depth_raw(height, width);
94  vpImage<vpRGBa> I_depth;
95 
96  g.open(settings);
97 
98  std::cout << "I_color: " << I_color.getWidth() << " " << I_color.getHeight() << std::endl;
99  std::cout << "I_depth_raw: " << I_depth_raw.getWidth() << " " << I_depth_raw.getHeight() << std::endl;
100 
101  g.acquire(reinterpret_cast<unsigned char *>(I_color.bitmap), reinterpret_cast<unsigned char *>(I_depth_raw.bitmap));
102 
103  std::cout << "Read camera parameters from Structure core device" << std::endl;
104  vpCameraParameters cam;
107 
108  std::cout << cam << std::endl;
109  std::cout << "poseEstimationMethod: " << poseEstimationMethod << std::endl;
110  std::cout << "tagFamily: " << tagFamily << std::endl;
111  std::cout << "nThreads : " << nThreads << std::endl;
112  std::cout << "Z aligned: " << align_frame << std::endl;
113 
114  vpImage<vpRGBa> I_color2 = I_color;
115  vpImage<float> depthMap;
116  vpImageConvert::createDepthHistogram(I_depth_raw, I_depth);
117 
118  vpDisplay *d1 = NULL;
119  vpDisplay *d2 = NULL;
120  vpDisplay *d3 = NULL;
121  if (! display_off) {
122 #ifdef VISP_HAVE_X11
123  d1 = new vpDisplayX(I_color, 100, 30, "Pose from Homography");
124  d2 = new vpDisplayX(I_color2, I_color.getWidth()+120, 30, "Pose from RGBD fusion");
125  d3 = new vpDisplayX(I_depth, 100, I_color.getHeight()+70, "Depth");
126 #elif defined(VISP_HAVE_GDI)
127  d1 = new vpDisplayGDI(I_color, 100, 30, "Pose from Homography");
128  d2 = new vpDisplayGDI(I_color2, I_color.getWidth()+120, 30, "Pose from RGBD fusion");
129  d3 = new vpDisplayGDI(I_depth, 100, I_color.getHeight()+70, "Depth");
130 #elif defined(VISP_HAVE_OPENCV)
131  d1 = new vpDisplayOpenCV(I_color, 100, 30, "Pose from Homography");
132  d2 = new vpDisplayOpenCV(I_color2, I_color.getWidth()+120, 30, "Pose from RGBD fusion");
133  d3 = new vpDisplayOpenCV(I_depth, 100, I_color.getHeight()+70, "Depth");
134 #endif
135  }
136 
138  vpDetectorAprilTag detector(tagFamily);
140 
142  detector.setAprilTagQuadDecimate(quad_decimate);
143  detector.setAprilTagPoseEstimationMethod(poseEstimationMethod);
144  detector.setAprilTagNbThreads(nThreads);
145  detector.setDisplayTag(display_tag, color_id < 0 ? vpColor::none : vpColor::getColor(color_id), thickness);
146  detector.setZAlignedWithCameraAxis(align_frame);
148  std::vector<double> time_vec;
149  for (;;) {
150  double t = vpTime::measureTimeMs();
151 
153  g.acquire(reinterpret_cast<unsigned char *>(I_color.bitmap), reinterpret_cast<unsigned char *>(I_depth_raw.bitmap));
155 
156  I_color2 = I_color;
157  vpImageConvert::convert(I_color, I);
158  vpImageConvert::createDepthHistogram(I_depth_raw, I_depth);
159 
160  depthMap.resize(I_depth_raw.getHeight(), I_depth_raw.getWidth());
161  #ifdef VISP_HAVE_OPENMP
162  #pragma omp parallel for
163  #endif
164  for (unsigned int i = 0; i < I_depth_raw.getHeight(); i++) {
165  for (unsigned int j = 0; j < I_depth_raw.getWidth(); j++) {
166  if (!vpMath::isNaN(I_depth_raw[i][j])) {
167  float Z = I_depth_raw[i][j] * 0.001; // Transform depth to meters.
168  depthMap[i][j] = Z;
169  } else {
170  depthMap[i][j] = 0;
171  }
172  }
173  }
174 
175  vpDisplay::display(I_color);
176  vpDisplay::display(I_color2);
177  vpDisplay::display(I_depth);
178 
179  std::vector<vpHomogeneousMatrix> cMo_vec;
180  detector.detect(I, tagSize, cam, cMo_vec);
181 
182  // Display camera pose for each tag
183  for (size_t i = 0; i < cMo_vec.size(); i++) {
184  vpDisplay::displayFrame(I_color, cMo_vec[i], cam, tagSize / 2, vpColor::none, 3);
185  }
186 
188  std::vector<std::vector<vpImagePoint> > tags_corners = detector.getPolygon();
189  std::vector<int> tags_id = detector.getTagsId();
190  std::map<int, double> tags_size;
191  tags_size[-1] = tagSize; // Default tag size
192  std::vector<std::vector<vpPoint> > tags_points3d = detector.getTagsPoints3D(tags_id, tags_size);
193  for (size_t i = 0; i < tags_corners.size(); i++) {
195  double confidence_index;
196  if (vpPose::computePlanarObjectPoseFromRGBD(depthMap, tags_corners[i], cam, tags_points3d[i], cMo, &confidence_index)) {
197  if (confidence_index > 0.5) {
198  vpDisplay::displayFrame(I_color2, cMo, cam, tagSize/2, vpColor::none, 3);
199  }
200  else if (confidence_index > 0.25) {
201  vpDisplay::displayFrame(I_color2, cMo, cam, tagSize/2, vpColor::orange, 3);
202  }
203  else {
204  vpDisplay::displayFrame(I_color2, cMo, cam, tagSize/2, vpColor::red, 3);
205  }
206  std::stringstream ss;
207  ss << "Tag id " << tags_id[i] << " confidence: " << confidence_index;
208  vpDisplay::displayText(I_color2, 35 + i*15, 20, ss.str(), vpColor::red);
209  }
210  }
212 
213  vpDisplay::displayText(I_color, 20, 20, "Pose from homography + VVS", vpColor::red);
214  vpDisplay::displayText(I_color2, 20, 20, "Pose from RGBD fusion", vpColor::red);
215  vpDisplay::displayText(I_color, 35, 20, "Click to quit.", vpColor::red);
216  t = vpTime::measureTimeMs() - t;
217  time_vec.push_back(t);
218 
219  std::stringstream ss;
220  ss << "Detection time: " << t << " ms for " << detector.getNbObjects() << " tags";
221  vpDisplay::displayText(I_color, 50, 20, ss.str(), vpColor::red);
222 
223  if (vpDisplay::getClick(I_color, false))
224  break;
225 
226  vpDisplay::flush(I_color);
227  vpDisplay::flush(I_color2);
228  vpDisplay::flush(I_depth);
229  }
230 
231  std::cout << "Benchmark loop processing time" << std::endl;
232  std::cout << "Mean / Median / Std: " << vpMath::getMean(time_vec) << " ms"
233  << " ; " << vpMath::getMedian(time_vec) << " ms"
234  << " ; " << vpMath::getStdev(time_vec) << " ms" << std::endl;
235 
236  if (! display_off) {
237  delete d1;
238  delete d2;
239  delete d3;
240  }
241 
242  } catch (const vpException &e) {
243  std::cerr << "Catch an exception: " << e.getMessage() << std::endl;
244  }
245 
246  return EXIT_SUCCESS;
247 #else
248  (void)argc;
249  (void)argv;
250 #ifndef VISP_HAVE_APRILTAG
251  std::cout << "Enable Apriltag support, configure and build ViSP to run this tutorial" << std::endl;
252 #else
253  std::cout << "Install Structure Core SDK, configure and build ViSP again to use this example" << std::endl;
254 #endif
255 #endif
256  return EXIT_SUCCESS;
257 }
Class that defines generic functionnalities for display.
Definition: vpDisplay.h:177
static bool getClick(const vpImage< unsigned char > &I, bool blocking=true)
void setAprilTagPoseEstimationMethod(const vpPoseEstimationMethod &poseEstimationMethod)
static void convert(const vpImage< unsigned char > &src, vpImage< vpRGBa > &dest)
void resize(unsigned int h, unsigned int w)
resize the image : Image initialization
Definition: vpImage.h:800
Implementation of an homogeneous matrix and operations on such kind of matrices.
AprilTag 36h11 pattern (recommended)
static double getMedian(const std::vector< double > &v)
Definition: vpMath.cpp:261
Display for windows using GDI (available on any windows 32 platform).
Definition: vpDisplayGDI.h:128
static double getStdev(const std::vector< double > &v, bool useBesselCorrection=false)
Definition: vpMath.cpp:291
static void displayText(const vpImage< unsigned char > &I, const vpImagePoint &ip, const std::string &s, const vpColor &color)
Use the X11 console to display images on unix-like OS. Thus to enable this class X11 should be instal...
Definition: vpDisplayX.h:134
static const vpColor none
Definition: vpColor.h:229
error that can be emited by ViSP classes.
Definition: vpException.h:71
std::vector< std::vector< vpPoint > > getTagsPoints3D(const std::vector< int > &tagsId, const std::map< int, double > &tagsSize) const
static bool computePlanarObjectPoseFromRGBD(const vpImage< float > &depthMap, const std::vector< vpImagePoint > &corners, const vpCameraParameters &colorIntrinsics, const std::vector< vpPoint > &point3d, vpHomogeneousMatrix &cMo, double *confidence_index=NULL)
Definition: vpPoseRGBD.cpp:254
size_t getNbObjects() const
const char * getMessage() const
Definition: vpException.cpp:90
bool open(const ST::CaptureSessionSettings &settings)
static void flush(const vpImage< unsigned char > &I)
VISP_EXPORT double measureTimeMs()
Definition: vpTime.cpp:126
static const vpColor red
Definition: vpColor.h:217
static const vpColor orange
Definition: vpColor.h:227
void setAprilTagQuadDecimate(float quadDecimate)
static double getMean(const std::vector< double > &v)
Definition: vpMath.cpp:241
void acquire(vpImage< unsigned char > &gray, bool undistorted=false, double *ts=NULL)
void setZAlignedWithCameraAxis(bool zAlignedWithCameraFrame)
static void display(const vpImage< unsigned char > &I)
The vpDisplayOpenCV allows to display image using the OpenCV library. Thus to enable this class OpenC...
Generic class defining intrinsic camera parameters.
void setAprilTagNbThreads(int nThreads)
static bool isNaN(double value)
Definition: vpMath.cpp:85
std::vector< int > getTagsId() const
unsigned int getHeight() const
Definition: vpImage.h:188
vpCameraParameters getCameraParameters(const vpOccipitalStructureStream stream_type, vpCameraParameters::vpCameraParametersProjType type=vpCameraParameters::perspectiveProjWithoutDistortion)
static void displayFrame(const vpImage< unsigned char > &I, const vpHomogeneousMatrix &cMo, const vpCameraParameters &cam, double size, const vpColor &color=vpColor::none, unsigned int thickness=1, const vpImagePoint &offset=vpImagePoint(0, 0))
void setDisplayTag(bool display, const vpColor &color=vpColor::none, unsigned int thickness=2)
static vpColor getColor(const unsigned int &i)
Definition: vpColor.h:310
unsigned int getWidth() const
Definition: vpImage.h:246
static void createDepthHistogram(const vpImage< uint16_t > &src_depth, vpImage< vpRGBa > &dest_rgba)
std::vector< std::vector< vpImagePoint > > & getPolygon()
bool detect(const vpImage< unsigned char > &I)