Example of tracking with vpGenericTracker on Castel.
#include <cstdlib>
#include <iostream>
#include <visp3/core/vpConfig.h>
#if (defined(VISP_HAVE_MODULE_MBT) && defined(VISP_HAVE_DISPLAY)) && \
(defined(VISP_HAVE_LAPACK) || defined(VISP_HAVE_EIGEN3) || defined(VISP_HAVE_OPENCV))
#include <visp3/core/vpDebug.h>
#include <visp3/core/vpHomogeneousMatrix.h>
#include <visp3/core/vpIoTools.h>
#include <visp3/core/vpMath.h>
#include <visp3/gui/vpDisplayD3D.h>
#include <visp3/gui/vpDisplayGDI.h>
#include <visp3/gui/vpDisplayGTK.h>
#include <visp3/gui/vpDisplayOpenCV.h>
#include <visp3/gui/vpDisplayX.h>
#include <visp3/io/vpImageIo.h>
#include <visp3/io/vpParseArgv.h>
#include <visp3/io/vpVideoReader.h>
#include <visp3/mbt/vpMbGenericTracker.h>
#define GETOPTARGS "x:X:m:M:i:n:dchfolwvpt:T:e:"
#define USE_SMALL_DATASET 1
#ifdef ENABLE_VISP_NAMESPACE
#endif
namespace
{
void usage(const char *name, const char *badparam)
{
fprintf(stdout, "\n\
Example of tracking with vpGenericTracker.\n\
\n\
SYNOPSIS\n\
%s [-i <test image path>] [-x <config file>] [-X <config file depth>]\n\
[-m <model name>] [-M <model name depth>] [-n <initialisation file base name>]\n\
[-f] [-c] [-d] [-h] [-o] [-w] [-l] [-v] [-p]\n\
[-t <tracker type>] [-T <tracker type>] [-e <last frame index>]\n",
name);
fprintf(stdout, "\n\
OPTIONS: \n\
-i <input image path> \n\
Set image input path.\n\
These images come from visp-images-x.y.z.tar.gz available \n\
on the ViSP website.\n\
Setting the VISP_INPUT_IMAGE_PATH environment\n\
variable produces the same behavior than using\n\
this option.\n\
\n\
-x <config file> \n\
Set the config file (the xml file) to use.\n\
The config file is used to specify the parameters of the tracker.\n\
\n\
-X <config file> \n\
Set the config file (the xml file) to use for the depth sensor.\n\
The config file is used to specify the parameters of the tracker.\n\
\n\
-m <model name> \n\
Specify the name of the file of the model.\n\
The model can either be a vrml model (.wrl) or a .cao file.\n\
\n\
-M <model name> \n\
Specify the name of the file of the model for the depth sensor.\n\
The model can either be a vrml model (.wrl) or a .cao file.\n\
\n\
-n <initialisation file base name> \n\
Base name of the initialisation file. The file will be 'base_name'.init .\n\
This base name is also used for the optional picture specifying where to \n\
click (a .ppm picture).\n\
\n\
-f \n\
Turn off the display of the the moving edges and Klt points. \n\
\n\
-d \n\
Turn off the display.\n\
\n\
-c\n\
Disable the mouse click. Useful to automate the \n\
execution of this program without human intervention.\n\
\n\
-o\n\
Use Ogre3D for visibility tests\n\
\n\
-w\n\
When Ogre3D is enable [-o] show Ogre3D configuration dialog that allows to set the renderer.\n\
\n\
-l\n\
Use the scanline for visibility tests.\n\
\n\
-v\n\
Compute covariance matrix.\n\
\n\
-p\n\
Compute gradient projection error.\n\
\n\
-t <tracker type>\n\
Set tracker type (<1 (Edge)>, <2 (KLT)>, <3 (both)>) for color sensor.\n\
\n\
-T <tracker type>\n\
Set tracker type (<4 (Depth normal)>, <8 (Depth dense)>, <12 (both)>) for depth sensor.\n\
\n\
-e <last frame index>\n\
Specify the index of the last frame. Once reached, the tracking is stopped.\n\
\n\
-h \n\
Print the help.\n\n");
if (badparam)
fprintf(stdout, "\nERROR: Bad parameter [%s]\n", badparam);
}
bool getOptions(int argc, const char **argv, std::string &ipath, std::string &configFile, std::string &configFile_depth,
std::string &modelFile, std::string &modelFile_depth, std::string &initFile, bool &displayFeatures,
bool &click_allowed, bool &display, bool &useOgre, bool &showOgreConfigDialog, bool &useScanline,
bool &computeCovariance, bool &projectionError, int &trackerType, int &tracker_type_depth,
int &lastFrame)
{
const char *optarg_;
int c;
switch (c) {
case 'i':
ipath = optarg_;
break;
case 'x':
configFile = optarg_;
break;
case 'X':
configFile_depth = optarg_;
break;
case 'm':
modelFile = optarg_;
break;
case 'M':
modelFile_depth = optarg_;
break;
case 'n':
initFile = optarg_;
break;
case 'f':
displayFeatures = false;
break;
case 'c':
click_allowed = false;
break;
case 'd':
display = false;
break;
case 'o':
useOgre = true;
break;
case 'l':
useScanline = true;
break;
case 'w':
showOgreConfigDialog = true;
break;
case 'v':
computeCovariance = true;
break;
case 'p':
projectionError = true;
break;
case 't':
trackerType = atoi(optarg_);
break;
case 'T':
tracker_type_depth = atoi(optarg_);
break;
case 'e':
lastFrame = atoi(optarg_);
break;
case 'h':
usage(argv[0], nullptr);
return false;
break;
default:
usage(argv[0], optarg_);
return false;
break;
}
}
if ((c == 1) || (c == -1)) {
usage(argv[0], nullptr);
std::cerr << "ERROR: " << std::endl;
std::cerr << " Bad argument " << optarg_ << std::endl << std::endl;
return false;
}
return true;
}
struct vpRealsenseIntrinsics_t
{
float ppx;
float ppy;
float fx;
float fy;
float coeffs[5];
};
void rs_deproject_pixel_to_point(float point[3], const vpRealsenseIntrinsics_t &intrin, const float pixel[2], float depth)
{
float x = (pixel[0] - intrin.ppx) / intrin.fx;
float y = (pixel[1] - intrin.ppy) / intrin.fy;
float r2 = x * x + y * y;
float f = 1 + intrin.coeffs[0] * r2 + intrin.coeffs[1] * r2 * r2 + intrin.coeffs[4] * r2 * r2 * r2;
float ux = x * f + 2 * intrin.coeffs[2] * x * y + intrin.coeffs[3] * (r2 + 2 * x * x);
float uy = y * f + 2 * intrin.coeffs[3] * x * y + intrin.coeffs[2] * (r2 + 2 * y * y);
x = ux;
y = uy;
point[0] = depth * x;
point[1] = depth * y;
point[2] = depth;
}
vpImage<uint16_t> &I_depth_raw, std::vector<vpColVector> &pointcloud,
unsigned int &pointcloud_width,
unsigned int &pointcloud_height)
{
#if VISP_HAVE_DATASET_VERSION >= 0x030600
std::string ext("png");
#else
std::string ext("pgm");
#endif
std::stringstream ss;
ss << input_directory << "/image_";
ss << std::setfill('0') << std::setw(4);
ss << cpt;
ss << ".";
ss << ext;
std::string filename_image = ss.str();
std::cerr << "Cannot read: " << filename_image << std::endl;
return false;
}
ss.str("");
ss << input_directory << "/depth_image_";
ss << std::setfill('0') << std::setw(4);
ss << cpt;
ss << ".bin";
std::string filename_depth = ss.str();
std::ifstream file_depth(filename_depth.c_str(), std::ios::in | std::ios::binary);
if (!file_depth.is_open()) {
return false;
}
unsigned int height = 0, width = 0;
I_depth_raw.
resize(height, width);
uint16_t depth_value = 0;
for (unsigned int i = 0; i < height; i++) {
for (unsigned int j = 0; j < width; j++) {
I_depth_raw[i][j] = depth_value;
}
}
pointcloud_width = width;
pointcloud_height = height;
pointcloud.
resize((
size_t)width * height);
const float depth_scale = 0.000124986647f;
vpRealsenseIntrinsics_t depth_intrinsic;
depth_intrinsic.ppx = 311.484558f;
depth_intrinsic.ppy = 246.283234f;
depth_intrinsic.fx = 476.053619f;
depth_intrinsic.fy = 476.053497f;
depth_intrinsic.coeffs[0] = 0.165056542f;
depth_intrinsic.coeffs[1] = -0.0508309528f;
depth_intrinsic.coeffs[2] = 0.00435937941f;
depth_intrinsic.coeffs[3] = 0.00541406544f;
depth_intrinsic.coeffs[4] = 0.250085592f;
for (unsigned int i = 0; i < height; i++) {
for (unsigned int j = 0; j < width; j++) {
float scaled_depth = I_depth_raw[i][j] * depth_scale;
float point[3];
float pixel[2] = { (float)j, (float)i };
rs_deproject_pixel_to_point(point, depth_intrinsic, pixel, scaled_depth);
data_3D[0] = point[0];
data_3D[1] = point[1];
data_3D[2] = point[2];
pointcloud[(size_t)(i * width + j)] = data_3D;
}
}
return true;
}
const std::string &
#if defined(VISP_HAVE_PUGIXML)
configFile
#endif
,
const std::string &
#if defined(VISP_HAVE_PUGIXML)
configFile_depth
#endif
)
{
#if defined(VISP_HAVE_PUGIXML)
dynamic_cast<vpMbGenericTracker *
>(tracker)->loadConfigFile(configFile, configFile_depth);
#else
#if defined(VISP_HAVE_MODULE_KLT) && defined(VISP_HAVE_OPENCV) && defined(HAVE_OPENCV_IMGPROC) && defined(HAVE_OPENCV_VIDEO)
#endif
dynamic_cast<vpMbGenericTracker *
>(tracker)->setDepthNormalPclPlaneEstimationRansacMaxIter(200);
dynamic_cast<vpMbGenericTracker *
>(tracker)->setDepthNormalPclPlaneEstimationRansacThreshold(0.001);
#endif
}
}
int main(int argc, const char **argv)
{
try {
std::string env_ipath;
std::string opt_ipath;
std::string ipath;
std::string opt_configFile;
std::string opt_configFile_depth;
std::string opt_modelFile;
std::string opt_modelFile_depth;
std::string opt_initFile;
std::string initFile;
bool displayFeatures = true;
bool opt_click_allowed = true;
bool opt_display = true;
bool useOgre = false;
bool showOgreConfigDialog = false;
bool useScanline = false;
bool computeCovariance = false;
bool projectionError = false;
#if defined(__mips__) || defined(__mips) || defined(mips) || defined(__MIPS__)
int opt_lastFrame = 5;
#else
int opt_lastFrame = -1;
#endif
if (!env_ipath.empty())
ipath = env_ipath;
if (!getOptions(argc, argv, opt_ipath, opt_configFile, opt_configFile_depth, opt_modelFile, opt_modelFile_depth,
opt_initFile, displayFeatures, opt_click_allowed, opt_display, useOgre, showOgreConfigDialog,
useScanline, computeCovariance, projectionError, trackerType_image, trackerType_depth,
opt_lastFrame)) {
return EXIT_FAILURE;
}
#if !(defined(VISP_HAVE_MODULE_KLT) && defined(VISP_HAVE_OPENCV) && defined(HAVE_OPENCV_IMGPROC) && defined(HAVE_OPENCV_VIDEO))
if (trackerType_image == 2 || trackerType_image == 3) {
std::cout << "KLT features cannot be used: ViSP is not built with "
"KLT module or OpenCV imgproc and video modules are not available."
<< std::endl;
return EXIT_SUCCESS;
}
#endif
if (opt_ipath.empty() && env_ipath.empty()) {
usage(argv[0], nullptr);
std::cerr << std::endl << "ERROR:" << std::endl;
std::cerr << " Use -i <visp image path> option or set VISP_INPUT_IMAGE_PATH " << std::endl
<< " environment variable to specify the location of the " << std::endl
<< " image path where test images are located." << std::endl
<< std::endl;
return EXIT_FAILURE;
}
std::cerr << "ViSP-images does not contain the folder: " << dir_path << "!" << std::endl;
return EXIT_SUCCESS;
}
std::string configFile, configFile_depth;
if (!opt_configFile.empty())
configFile = opt_configFile;
else
configFile =
if (!opt_configFile_depth.empty())
configFile_depth = opt_configFile_depth;
else
configFile_depth =
std::string modelFile, modelFile_depth;
if (!opt_modelFile.empty())
modelFile = opt_modelFile;
else {
#if defined(VISP_HAVE_COIN3D) && (COIN_MAJOR_VERSION == 2 || COIN_MAJOR_VERSION == 3 || COIN_MAJOR_VERSION == 4)
modelFile =
#else
#endif
}
if (!opt_modelFile_depth.empty())
modelFile_depth = opt_modelFile_depth;
else
modelFile_depth =
std::string vrml_ext = ".wrl";
bool use_vrml =
(modelFile.compare(modelFile.length() - vrml_ext.length(), vrml_ext.length(), vrml_ext) == 0) ||
(modelFile_depth.compare(modelFile_depth.length() - vrml_ext.length(), vrml_ext.length(), vrml_ext) == 0);
if (use_vrml) {
#if defined(VISP_HAVE_COIN3D) && (COIN_MAJOR_VERSION == 2 || COIN_MAJOR_VERSION == 3 || COIN_MAJOR_VERSION == 4)
std::cout << "use_vrml: " << use_vrml << std::endl;
#else
std::cerr << "Error: vrml model file is only supported if ViSP is "
"build with Coin3D 3rd party"
<< std::endl;
return EXIT_FAILURE;
#endif
}
if (!opt_initFile.empty())
initFile = opt_initFile;
else
std::vector<vpColVector> pointcloud;
unsigned int pointcloud_width, pointcloud_height;
if (!read_data(0, ipath, I, I_depth_raw, pointcloud, pointcloud_width, pointcloud_height)) {
std::cerr << "Cannot open sequence: " << ipath << std::endl;
return EXIT_FAILURE;
}
#if defined(VISP_HAVE_X11)
vpDisplayX display1, display2;
#elif defined(VISP_HAVE_GDI)
#elif defined(HAVE_OPENCV_HIGHGUI)
#elif defined(VISP_HAVE_D3D9)
#elif defined(VISP_HAVE_GTK)
#else
opt_display = false;
#endif
if (opt_display) {
#if defined(VISP_HAVE_DISPLAY)
display1.init(I, 100, 100, "Test tracking (Left)");
"Test tracking (Right)");
#endif
}
std::vector<int> trackerTypes(2);
trackerTypes[0] = trackerType_image;
trackerTypes[1] = trackerType_depth;
loadConfiguration(tracker, configFile, configFile_depth);
std::string depth_M_color_filename =
{
std::ifstream depth_M_color_file(depth_M_color_filename.c_str());
depth_M_color.
load(depth_M_color_file);
std::map<std::string, vpHomogeneousMatrix> mapOfCameraTransformationMatrices;
mapOfCameraTransformationMatrices["Camera2"] = depth_M_color;
dynamic_cast<vpMbGenericTracker *
>(tracker)->setCameraTransformationMatrix(mapOfCameraTransformationMatrices);
}
if (useOgre)
if (opt_display && opt_click_allowed) {
}
}
if (opt_display && opt_click_allowed) {
std::map<std::string, const vpImage<unsigned char> *> mapOfImages;
mapOfImages["Camera1"] = &I;
mapOfImages["Camera2"] = &I_depth;
std::map<std::string, std::string> mapOfInitFiles;
mapOfInitFiles["Camera1"] = initFile;
dynamic_cast<vpMbGenericTracker *
>(tracker)->initClick(mapOfImages, mapOfInitFiles,
true);
}
else {
vpHomogeneousMatrix c1Moi(0.06846423368, 0.09062570884, 0.3401096693, -2.671882598, 0.1174275908, -0.6011935263);
vpHomogeneousMatrix c2Moi(0.04431452054, 0.09294637757, 0.3357760654, -2.677922443, 0.121297639, -0.6028463357);
}
{
std::map<std::string, const vpImage<unsigned char> *> mapOfImages;
mapOfImages["Camera1"] = &I;
std::map<std::string, const std::vector<vpColVector> *> mapOfPointclouds;
mapOfPointclouds["Camera2"] = &pointcloud;
std::map<std::string, unsigned int> mapOfWidths, mapOfHeights;
mapOfWidths["Camera2"] = pointcloud_width;
mapOfHeights["Camera2"] = pointcloud_height;
dynamic_cast<vpMbGenericTracker *
>(tracker)->track(mapOfImages, mapOfPointclouds, mapOfWidths, mapOfHeights);
}
if (opt_display) {
}
bool quit = false, click = false;
unsigned int frame_index = 0;
std::vector<double> time_vec;
while (read_data(frame_index, ipath, I, I_depth_raw, pointcloud, pointcloud_width, pointcloud_height) && !quit &&
(opt_lastFrame > 0 ? (int)frame_index <= opt_lastFrame : true)) {
if (opt_display) {
std::stringstream ss;
ss << "Num frame: " << frame_index;
}
if (frame_index == 10) {
std::cout << "----------Test reset tracker----------" << std::endl;
if (opt_display) {
}
loadConfiguration(tracker, configFile, configFile_depth);
}
#if USE_SMALL_DATASET
if (frame_index == 20) {
c1Mo.
buildFrom(0.07734634051, 0.08993639906, 0.342344402, -2.708409543, 0.0669276477, -0.3798958303);
c2Mo.
buildFrom(0.05319520317, 0.09223511976, 0.3380095812, -2.71438192, 0.07141055397, -0.3810081638);
#else
if (frame_index == 50) {
c1Mo.
buildFrom(0.09280663035, 0.09277655672, 0.330415149, -2.724431817, 0.0293932671, 0.02027966377);
c2Mo.
buildFrom(0.06865933578, 0.09494713501, 0.3260555142, -2.730027451, 0.03498390135, 0.01989831338);
#endif
std::cout << "Test set pose" << std::endl;
}
#if USE_SMALL_DATASET
if (frame_index < 15 || frame_index >= 20) {
#else
if (frame_index < 30 || frame_index >= 50) {
#endif
std::map<std::string, const vpImage<unsigned char> *> mapOfImages;
mapOfImages["Camera1"] = &I;
std::map<std::string, const std::vector<vpColVector> *> mapOfPointclouds;
mapOfPointclouds["Camera2"] = &pointcloud;
std::map<std::string, unsigned int> mapOfWidths, mapOfHeights;
mapOfWidths["Camera2"] = pointcloud_width;
mapOfHeights["Camera2"] = pointcloud_height;
dynamic_cast<vpMbGenericTracker *
>(tracker)->track(mapOfImages, mapOfPointclouds, mapOfWidths, mapOfHeights);
time_vec.push_back(t);
if (opt_display) {
std::stringstream ss;
ss << "Computation time: " << t << " ms";
ss.str("");
{
std::stringstream ss;
ss <<
"Features: edges " <<
dynamic_cast<vpMbGenericTracker *
>(tracker)->getNbFeaturesEdge() <<
", klt "
}
}
}
if (opt_click_allowed && opt_display) {
switch (button) {
quit = !click;
break;
click = !click;
break;
default:
break;
}
}
}
if (computeCovariance) {
std::cout <<
"Covariance matrix: \n" << tracker->
getCovarianceMatrix() << std::endl << std::endl;
}
if (projectionError) {
std::cout <<
"Projection error: " << tracker->
getProjectionError() << std::endl << std::endl;
}
if (opt_display) {
}
frame_index++;
}
std::cout << "\nFinal poses, c1Mo:\n" << c1Mo << "\nc2Mo:\n" << c2Mo << std::endl;
<< std::endl;
if (opt_click_allowed && !quit) {
}
delete tracker;
tracker = nullptr;
return EXIT_SUCCESS;
}
std::cout << "Catch an exception: " << e << std::endl;
return EXIT_FAILURE;
}
}
#elif !(defined(VISP_HAVE_MODULE_MBT) && defined(VISP_HAVE_DISPLAY))
int main()
{
std::cout << "Cannot run this example: visp_mbt, visp_gui modules are required." << std::endl;
return EXIT_SUCCESS;
}
#else
int main()
{
std::cout << "Cannot run this example: install Lapack, Eigen3 or OpenCV" << std::endl;
return EXIT_SUCCESS;
}
#endif
unsigned int getRows() const
Generic class defining intrinsic camera parameters.
void initPersProjWithoutDistortion(double px, double py, double u0, double v0)
Implementation of column vector and the associated operations.
static const vpColor darkRed
Display for windows using Direct3D 3rd party. Thus to enable this class Direct3D should be installed....
Display for windows using GDI (available on any windows 32 platform).
The vpDisplayGTK allows to display image using the GTK 3rd party library. Thus to enable this class G...
The vpDisplayOpenCV allows to display image using the OpenCV library. Thus to enable this class OpenC...
static bool getClick(const vpImage< unsigned char > &I, bool blocking=true)
virtual void setDownScalingFactor(unsigned int scale)
static void display(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), const std::string &frameName="", const vpColor &textColor=vpColor::black, const vpImagePoint &textOffset=vpImagePoint(15, 15))
static void flush(const vpImage< unsigned char > &I)
unsigned int getDownScalingFactor()
static void displayText(const vpImage< unsigned char > &I, const vpImagePoint &ip, const std::string &s, const vpColor &color)
error that can be emitted by ViSP classes.
Implementation of an homogeneous matrix and operations on such kind of matrices.
void load(std::ifstream &f)
vpHomogeneousMatrix & buildFrom(const vpTranslationVector &t, const vpRotationMatrix &R)
static void createDepthHistogram(const vpImage< uint16_t > &src_depth, vpImage< vpRGBa > &dest_rgba)
static void read(vpImage< unsigned char > &I, const std::string &filename, int backend=IO_DEFAULT_BACKEND)
unsigned int getWidth() const
void resize(unsigned int h, unsigned int w)
resize the image : Image initialization
unsigned int getHeight() const
Wrapper for the KLT (Kanade-Lucas-Tomasi) feature tracker implemented in OpenCV. Thus to enable this ...
void setBlockSize(int blockSize)
void setQuality(double qualityLevel)
void setHarrisFreeParameter(double harris_k)
void setMaxFeatures(int maxCount)
void setMinDistance(double minDistance)
void setWindowSize(int winSize)
void setPyramidLevels(int pyrMaxLevel)
static double rad(double deg)
static double getMedian(const std::vector< double > &v)
static double getStdev(const std::vector< double > &v, bool useBesselCorrection=false)
static double getMean(const std::vector< double > &v)
Real-time 6D object pose tracking using its CAD model.
Main methods for a model-based tracker.
virtual void resetTracker()=0
virtual void setOgreShowConfigDialog(bool showConfigDialog)
virtual void setDisplayFeatures(bool displayF)
virtual vpColVector getError() const =0
virtual void setAngleDisappear(const double &a)
virtual void setCovarianceComputation(const bool &flag)
virtual void setScanLineVisibilityTest(const bool &v)
virtual void setOgreVisibilityTest(const bool &v)
virtual vpMatrix getCovarianceMatrix() const
virtual void setNearClippingDistance(const double &dist)
virtual void setFarClippingDistance(const double &dist)
virtual double getProjectionError() const
virtual void setProjectionErrorComputation(const bool &flag)
virtual void setClipping(const unsigned int &flags)
virtual void setAngleAppear(const double &a)
virtual unsigned int getClipping() const
@ ROBUST_FEATURE_ESTIMATION
Robust scheme to estimate the normal of the plane.
void setMu1(const double &mu_1)
void setRange(const unsigned int &range)
void setLikelihoodThresholdType(const vpLikelihoodThresholdType likelihood_threshold_type)
void setMaskNumber(const unsigned int &mask_number)
void setThreshold(const double &threshold)
void setSampleStep(const double &sample_step)
void setMaskSize(const unsigned int &mask_size)
void setMu2(const double &mu_2)
static bool parse(int *argcPtr, const char **argv, vpArgvInfo *argTable, int flags)
VISP_EXPORT double measureTimeMs()