Visual Servoing Platform  version 3.5.1 under development (2022-08-10)
tutorial-detection-object-mbt2.cpp
1 #include <visp3/core/vpConfig.h>
3 #include <visp3/core/vpIoTools.h>
4 #include <visp3/gui/vpDisplayGDI.h>
5 #include <visp3/gui/vpDisplayOpenCV.h>
6 #include <visp3/gui/vpDisplayX.h>
7 #include <visp3/io/vpVideoReader.h>
8 #include <visp3/mbt/vpMbGenericTracker.h>
9 #include <visp3/vision/vpKeyPoint.h>
10 
11 #if (VISP_HAVE_OPENCV_VERSION >= 0x020400)
12 void learnCube(const vpImage<unsigned char> &I, vpMbGenericTracker &tracker, vpKeyPoint &keypoint_learning, int id)
13 {
15  std::vector<cv::KeyPoint> trainKeyPoints;
16  double elapsedTime;
17  keypoint_learning.detect(I, trainKeyPoints, elapsedTime);
19 
21  std::vector<vpPolygon> polygons;
22  std::vector<std::vector<vpPoint> > roisPt;
23  std::pair<std::vector<vpPolygon>, std::vector<std::vector<vpPoint> > > pair = tracker.getPolygonFaces();
24  polygons = pair.first;
25  roisPt = pair.second;
26 
27  std::vector<cv::Point3f> points3f;
29  tracker.getPose(cMo);
31  tracker.getCameraParameters(cam);
32  vpKeyPoint::compute3DForPointsInPolygons(cMo, cam, trainKeyPoints, polygons, roisPt, points3f);
34 
36  keypoint_learning.buildReference(I, trainKeyPoints, points3f, true, id);
38 
40  for (std::vector<cv::KeyPoint>::const_iterator it = trainKeyPoints.begin(); it != trainKeyPoints.end(); ++it) {
41  vpDisplay::displayCross(I, (int)it->pt.y, (int)it->pt.x, 4, vpColor::red);
42  }
44 }
45 #endif
46 
47 int main(int argc, char **argv)
48 {
49 #if (VISP_HAVE_OPENCV_VERSION >= 0x020400)
51  try {
52  std::string videoname = "cube.mp4";
53 
54  for (int i = 0; i < argc; i++) {
55  if (std::string(argv[i]) == "--name")
56  videoname = std::string(argv[i + 1]);
57  else if (std::string(argv[i]) == "--help" || std::string(argv[i]) == "-h") {
58  std::cout << "\nUsage: " << argv[0] << " [--name <video name>] [--help] [-h]\n" << std::endl;
59  return 0;
60  }
61  }
62  std::string parentname = vpIoTools::getParent(videoname);
63  std::string objectname = vpIoTools::getNameWE(videoname);
64 
65  if (!parentname.empty())
66  objectname = parentname + "/" + objectname;
67 
68  std::cout << "Video name: " << videoname << std::endl;
69  std::cout << "Tracker requested config files: " << objectname << ".[init,"
70  << "xml,"
71  << "cao or wrl]" << std::endl;
72  std::cout << "Tracker optional config files: " << objectname << ".[ppm]" << std::endl;
73 
77 
79  bool usexml = false;
80  if (vpIoTools::checkFilename(objectname + ".xml")) {
81  tracker.loadConfigFile(objectname + ".xml");
82  tracker.getCameraParameters(cam);
83  usexml = true;
84  }
85  if (!usexml) {
86  vpMe me;
87  me.setMaskSize(5);
88  me.setMaskNumber(180);
89  me.setRange(7);
90  me.setThreshold(5000);
91  me.setMu1(0.5);
92  me.setMu2(0.5);
93  me.setSampleStep(4);
94  me.setNbTotalSample(250);
95  tracker.setMovingEdge(me);
96  cam.initPersProjWithoutDistortion(547, 542, 339, 235);
97  tracker.setCameraParameters(cam);
98  tracker.setAngleAppear(vpMath::rad(89));
99  tracker.setAngleDisappear(vpMath::rad(89));
100  tracker.setNearClippingDistance(0.01);
101  tracker.setFarClippingDistance(10.0);
103  }
104 
105  tracker.setOgreVisibilityTest(false);
106  if (vpIoTools::checkFilename(objectname + ".cao"))
107  tracker.loadModel(objectname + ".cao");
108  else if (vpIoTools::checkFilename(objectname + ".wrl"))
109  tracker.loadModel(objectname + ".wrl");
110  tracker.setDisplayFeatures(true);
112 
114  vpKeyPoint keypoint_learning("ORB", "ORB", "BruteForce-Hamming");
115 #if (VISP_HAVE_OPENCV_VERSION < 0x030000)
116  keypoint_learning.setDetectorParameter("ORB", "nLevels", 1);
117 #else
118  cv::Ptr<cv::ORB> orb_learning = keypoint_learning.getDetector("ORB").dynamicCast<cv::ORB>();
119  if (orb_learning) {
120  orb_learning->setNLevels(1);
121  }
122 #endif
124 
125 #if defined(VISP_HAVE_X11)
126  vpDisplayX display;
127 #elif defined(VISP_HAVE_GDI)
128  vpDisplayGDI display;
129 #elif defined(VISP_HAVE_OPENCV)
130  vpDisplayOpenCV display;
131 #else
132  std::cout << "No image viewer is available..." << std::endl;
133  return 0;
134 #endif
135 
136  /*
137  * Start the part of the code dedicated to object learning from 3 images
138  */
139  std::string imageName[] = {"cube0001.png", "cube0150.png", "cube0200.png"};
140  vpHomogeneousMatrix initPoseTab[] = {
141  vpHomogeneousMatrix(0.02143385294, 0.1098083886, 0.5127439561, 2.087159614, 1.141775176, -0.4701291124),
142  vpHomogeneousMatrix(0.02651282185, -0.03713587374, 0.6873765919, 2.314744454, 0.3492296488, -0.1226054828),
143  vpHomogeneousMatrix(0.02965448956, -0.07283091786, 0.7253526051, 2.300529617, -0.4286674806, 0.1788761025)};
144  for (int i = 0; i < 3; i++) {
145  vpImageIo::read(I, imageName[i]);
146  if (i == 0) {
147  display.init(I, 10, 10);
148  }
149  std::stringstream title;
150  title << "Learning cube on image: " << imageName[i];
151  vpDisplay::setTitle(I, title.str().c_str());
152 
154 
156  tracker.setPose(I, initPoseTab[i]);
158 
160  tracker.track(I);
162 
164  tracker.getPose(cMo);
165  tracker.display(I, cMo, cam, vpColor::red);
167 
169  learnCube(I, tracker, keypoint_learning, i);
171 
172  vpDisplay::displayText(I, 10, 10, "Learning step: keypoints are detected on visible cube faces", vpColor::red);
173  if (i < 2) {
174  vpDisplay::displayText(I, 30, 10, "Click to continue the learning...", vpColor::red);
175  } else {
176  vpDisplay::displayText(I, 30, 10, "Click to continue with the detection...", vpColor::red);
177  }
178 
179  vpDisplay::flush(I);
180  vpDisplay::getClick(I, true);
181  }
182 
184  keypoint_learning.saveLearningData("cube_learning_data.bin", true);
186 
187  /*
188  * Start the part of the code dedicated to detection and localization
189  */
191  vpKeyPoint keypoint_detection("ORB", "ORB", "BruteForce-Hamming");
192 #if (VISP_HAVE_OPENCV_VERSION < 0x030000)
193  keypoint_detection.setDetectorParameter("ORB", "nLevels", 1);
194 #else
195  cv::Ptr<cv::ORB> orb_detector = keypoint_detection.getDetector("ORB").dynamicCast<cv::ORB>();
196  orb_detector = keypoint_detection.getDetector("ORB").dynamicCast<cv::ORB>();
197  if (orb_detector) {
198  orb_detector->setNLevels(1);
199  }
200 #endif
202 
204  keypoint_detection.loadLearningData("cube_learning_data.bin", true);
206 
208  vpImage<unsigned char> IMatching;
209  keypoint_detection.createImageMatching(I, IMatching);
211 
212  vpVideoReader g;
213  g.setFileName(videoname);
214  g.open(I);
215 
216 #if defined VISP_HAVE_X11
217  vpDisplayX display2;
218 #elif defined VISP_HAVE_GTK
219  vpDisplayGTK display2;
220 #elif defined VISP_HAVE_GDI
221  vpDisplayGDI display2;
222 #else
223  vpDisplayOpenCV display2;
224 #endif
225  display2.init(IMatching, 50, 50, "Display matching between learned and current images");
226  vpDisplay::setTitle(I, "Cube detection and localization");
227 
228  double error;
229  bool click_done = false;
230 
231  while (!g.end()) {
232  g.acquire(I);
234 
236  keypoint_detection.insertImageMatching(I, IMatching);
238 
239  vpDisplay::display(IMatching);
240  vpDisplay::displayText(I, 10, 10, "Detection and localization in process...", vpColor::red);
241 
242  double elapsedTime;
244  if (keypoint_detection.matchPoint(I, cam, cMo, error, elapsedTime)) {
246 
248  tracker.setPose(I, cMo);
250 
252  tracker.display(I, cMo, cam, vpColor::red, 2);
253  vpDisplay::displayFrame(I, cMo, cam, 0.05, vpColor::none, 3);
255 
256  keypoint_detection.displayMatching(I, IMatching);
257 
259  std::vector<vpImagePoint> ransacInliers = keypoint_detection.getRansacInliers();
260  std::vector<vpImagePoint> ransacOutliers = keypoint_detection.getRansacOutliers();
262 
264  for (std::vector<vpImagePoint>::const_iterator it = ransacInliers.begin(); it != ransacInliers.end(); ++it) {
266  vpImagePoint imPt(*it);
267  imPt.set_u(imPt.get_u() + I.getWidth());
268  imPt.set_v(imPt.get_v() + I.getHeight());
269  vpDisplay::displayCircle(IMatching, imPt, 4, vpColor::green);
270  }
272 
274  for (std::vector<vpImagePoint>::const_iterator it = ransacOutliers.begin(); it != ransacOutliers.end(); ++it) {
276  vpImagePoint imPt(*it);
277  imPt.set_u(imPt.get_u() + I.getWidth());
278  imPt.set_v(imPt.get_v() + I.getHeight());
279  vpDisplay::displayCircle(IMatching, imPt, 4, vpColor::red);
280  }
282 
284  keypoint_detection.displayMatching(I, IMatching);
286 
288  vpCameraParameters cam2;
289  cam2.initPersProjWithoutDistortion(cam.get_px(), cam.get_py(), cam.get_u0() + I.getWidth(),
290  cam.get_v0() + I.getHeight());
291  tracker.setCameraParameters(cam2);
292  tracker.setPose(IMatching, cMo);
293  tracker.display(IMatching, cMo, cam2, vpColor::red, 2);
294  vpDisplay::displayFrame(IMatching, cMo, cam2, 0.05, vpColor::none, 3);
296  }
297 
298  vpDisplay::flush(I);
299  vpDisplay::displayText(IMatching, 30, 10, "A click to exit.", vpColor::red);
300  vpDisplay::flush(IMatching);
301  if (vpDisplay::getClick(I, false)) {
302  click_done = true;
303  break;
304  }
305  if (vpDisplay::getClick(IMatching, false)) {
306  click_done = true;
307  break;
308  }
309  }
310 
311  if (!click_done)
312  vpDisplay::getClick(IMatching);
313  } catch (const vpException &e) {
314  std::cout << "Catch an exception: " << e << std::endl;
315  }
316 #else
317  (void)argc;
318  (void)argv;
319  std::cout << "Install OpenCV and rebuild ViSP to use this example." << std::endl;
320 #endif
321 
322  return 0;
323 }
Generic class defining intrinsic camera parameters.
void initPersProjWithoutDistortion(double px, double py, double u0, double v0)
static const vpColor red
Definition: vpColor.h:217
static const vpColor none
Definition: vpColor.h:229
static const vpColor green
Definition: vpColor.h:220
Display for windows using GDI (available on any windows 32 platform).
Definition: vpDisplayGDI.h:129
The vpDisplayGTK allows to display image using the GTK 3rd party library. Thus to enable this class G...
Definition: vpDisplayGTK.h:135
The vpDisplayOpenCV allows to display image using the OpenCV library. Thus to enable this class OpenC...
Use the X11 console to display images on unix-like OS. Thus to enable this class X11 should be instal...
Definition: vpDisplayX.h:135
void init(vpImage< unsigned char > &I, int win_x=-1, int win_y=-1, const std::string &win_title="")
static bool getClick(const vpImage< unsigned char > &I, bool blocking=true)
static void display(const vpImage< unsigned char > &I)
static void displayCross(const vpImage< unsigned char > &I, const vpImagePoint &ip, unsigned int size, const vpColor &color, unsigned int thickness=1)
static void setTitle(const vpImage< unsigned char > &I, const std::string &windowtitle)
static void flush(const vpImage< unsigned char > &I)
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))
static void displayText(const vpImage< unsigned char > &I, const vpImagePoint &ip, const std::string &s, const vpColor &color)
static void displayCircle(const vpImage< unsigned char > &I, const vpImagePoint &center, unsigned int radius, const vpColor &color, bool fill=false, unsigned int thickness=1)
error that can be emited by ViSP classes.
Definition: vpException.h:72
Implementation of an homogeneous matrix and operations on such kind of matrices.
static void read(vpImage< unsigned char > &I, const std::string &filename, int backend=IO_DEFAULT_BACKEND)
Definition: vpImageIo.cpp:148
Class that defines a 2D point in an image. This class is useful for image processing and stores only ...
Definition: vpImagePoint.h:89
unsigned int getWidth() const
Definition: vpImage.h:246
unsigned int getHeight() const
Definition: vpImage.h:188
static bool checkFilename(const std::string &filename)
Definition: vpIoTools.cpp:800
static std::string getNameWE(const std::string &pathname)
Definition: vpIoTools.cpp:1536
static std::string getParent(const std::string &pathname)
Definition: vpIoTools.cpp:1620
Class that allows keypoints detection (and descriptors extraction) and matching thanks to OpenCV libr...
Definition: vpKeyPoint.h:223
void detect(const vpImage< unsigned char > &I, std::vector< cv::KeyPoint > &keyPoints, const vpRect &rectangle=vpRect())
static void compute3DForPointsInPolygons(const vpHomogeneousMatrix &cMo, const vpCameraParameters &cam, std::vector< cv::KeyPoint > &candidates, const std::vector< vpPolygon > &polygons, const std::vector< std::vector< vpPoint > > &roisPt, std::vector< cv::Point3f > &points, cv::Mat *descriptors=NULL)
Definition: vpKeyPoint.cpp:634
void saveLearningData(const std::string &filename, bool binaryMode=false, bool saveTrainingImages=true)
unsigned int buildReference(const vpImage< unsigned char > &I)
Definition: vpKeyPoint.cpp:237
cv::Ptr< cv::FeatureDetector > getDetector(const vpFeatureDetectorType &type) const
Definition: vpKeyPoint.h:491
static double rad(double deg)
Definition: vpMath.h:117
Real-time 6D object pose tracking using its CAD model.
virtual void setCameraParameters(const vpCameraParameters &camera)
virtual void getPose(vpHomogeneousMatrix &cMo) const
virtual void setDisplayFeatures(bool displayF)
virtual void setAngleAppear(const double &a)
virtual void setNearClippingDistance(const double &dist)
virtual void getCameraParameters(vpCameraParameters &camera) const
virtual void setAngleDisappear(const double &a)
virtual void setMovingEdge(const vpMe &me)
virtual std::pair< std::vector< vpPolygon >, std::vector< std::vector< vpPoint > > > getPolygonFaces(bool orderPolygons=true, bool useVisibility=true, bool clipPolygon=false)
virtual void getClipping(unsigned int &clippingFlag1, unsigned int &clippingFlag2) const
virtual void setFarClippingDistance(const double &dist)
virtual void setClipping(const unsigned int &flags)
virtual void loadConfigFile(const std::string &configFile, bool verbose=true)
virtual void setPose(const vpImage< unsigned char > &I, const vpHomogeneousMatrix &cdMo)
virtual void setOgreVisibilityTest(const bool &v)
virtual void loadModel(const std::string &modelFile, bool verbose=false, const vpHomogeneousMatrix &T=vpHomogeneousMatrix())
virtual void display(const vpImage< unsigned char > &I, const vpHomogeneousMatrix &cMo, const vpCameraParameters &cam, const vpColor &col, unsigned int thickness=1, bool displayFullModel=false)
virtual void track(const vpImage< unsigned char > &I)
Definition: vpMe.h:61
void setMu1(const double &mu_1)
Definition: vpMe.h:241
void setSampleStep(const double &s)
Definition: vpMe.h:278
void setRange(const unsigned int &r)
Definition: vpMe.h:271
void setMaskSize(const unsigned int &a)
Definition: vpMe.cpp:459
void setNbTotalSample(const int &nb)
Definition: vpMe.h:255
void setMu2(const double &mu_2)
Definition: vpMe.h:248
void setMaskNumber(const unsigned int &a)
Definition: vpMe.cpp:452
void setThreshold(const double &t)
Definition: vpMe.h:300
Class that enables to manipulate easily a video file or a sequence of images. As it inherits from the...
void acquire(vpImage< vpRGBa > &I)
void open(vpImage< vpRGBa > &I)
void setFileName(const std::string &filename)