face_detection.cpp

Go to the documentation of this file.

/****************************************************************
 *
 * Copyright (c) 2010
 *
 * Fraunhofer Institute for Manufacturing Engineering   
 * and Automation (IPA)
 *
 * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 *
 * Project name: care-o-bot
 * ROS stack name: cob_vision
 * ROS package name: cob_people_detection
 * Description:
 *                                                              
 * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 *                      
 * Author: Jan Fischer, email:jan.fischer@ipa.fhg.de
 * Author: Richard Bormann, email: richard.bormann@ipa.fhg.de
 * Supervised by: 
 *
 * Date of creation: 03/2011
 * ToDo:
 *
 * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 *     * Redistributions of source code must retain the above copyright
 *       notice, this list of conditions and the following disclaimer.
 *     * Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in the
 *       documentation and/or other materials provided with the distribution.
 *     * Neither the name of the Fraunhofer Institute for Manufacturing 
 *       Engineering and Automation (IPA) nor the names of its
 *       contributors may be used to endorse or promote products derived from
 *       this software without specific prior written permission.
 *
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License LGPL as 
 * published by the Free Software Foundation, either version 3 of the 
 * License, or (at your option) any later version.
 * 
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Lesser General Public License LGPL for more details.
 * 
 * You should have received a copy of the GNU Lesser General Public 
 * License LGPL along with this program. 
 * If not, see <http://www.gnu.org/licenses/>.
 *
 ****************************************************************/

#include "cob_people_detection/face_detection.h"
#include <pluginlib/class_list_macros.h>

#ifdef __LINUX__
#include "cob_vision_utils/GlobalDefines.h"
#include "cob_vision_utils/VisionUtils.h"
#else
#include "cob_common/cob_vision_utils/common/include/cob_vision_utils/GlobalDefines.h"
#include "cob_common/cob_vision_utils/common/include/cob_vision_utils/VisionUtils.h"
#endif

#include <opencv/highgui.h>

PLUGINLIB_DECLARE_CLASS(ipa_PeopleDetector, CobFaceDetectionNodelet, ipa_PeopleDetector::CobFaceDetectionNodelet, nodelet::Nodelet)
;

using namespace ipa_PeopleDetector;

// Prevent deleting memory twice, when using smart pointer
void voidDeleter(sensor_msgs::PointCloud2* const )
{
}

//####################
//#### node class ####
CobFaceDetectionNodelet::CobFaceDetectionNodelet()
{
        it_ = 0;
        sync_pointcloud_ = 0;
        people_detector_ = 0;
        train_continuous_server_ = 0;
        train_capture_sample_server_ = 0;
        recognize_server_ = 0;
        load_server_ = 0;
        save_server_ = 0;
        show_server_ = 0;
        occupied_by_action_ = false;
        recognize_server_running_ = false;
        train_continuous_server_running_ = false;
        number_training_images_captured_ = 0;
        do_recognition_ = true;
        display_ = false;
        directory_ = ros::package::getPath("cob_people_detection") + "/common/files/windows/"; // can be changed by a parameter
}

CobFaceDetectionNodelet::~CobFaceDetectionNodelet()
{
        if (people_detector_ != 0)
                delete people_detector_;
        if (train_continuous_server_ != 0)
                delete train_continuous_server_;
        if (train_capture_sample_server_ != 0)
                delete train_capture_sample_server_;
        if (recognize_server_ != 0)
                delete recognize_server_;
        if (load_server_ != 0)
                delete load_server_;
        if (save_server_ != 0)
                delete save_server_;
        if (show_server_ != 0)
                delete show_server_;
        if (it_ != 0)
                delete it_;
        if (sync_pointcloud_ != 0)
                delete sync_pointcloud_;
}

void CobFaceDetectionNodelet::onInit()
{
        sync_pointcloud_ = new message_filters::Synchronizer<message_filters::sync_policies::ApproximateTime<sensor_msgs::PointCloud2, sensor_msgs::Image> >(2);
        node_handle_ = getNodeHandle();
        it_ = new image_transport::ImageTransport(node_handle_);
        face_position_publisher_ = node_handle_.advertise<cob_perception_msgs::DetectionArray>("face_position_array", 1);
        face_detection_image_pub_ = it_->advertise("face_detection_image", 1);

        recognize_server_ = new RecognizeServer(node_handle_, "recognize_server", boost::bind(&CobFaceDetectionNodelet::recognizeServerCallback, this, _1), false);
        recognize_server_->start();

        recognize_service_server_ = node_handle_.advertiseService("recognize_service_server", &CobFaceDetectionNodelet::recognizeServiceServerCallback, this);

        train_continuous_server_ = new TrainContinuousServer(node_handle_, "train_continuous_server", boost::bind(&CobFaceDetectionNodelet::trainContinuousServerCallback, this, _1),
                        false);
        train_continuous_server_->start();

        train_capture_sample_server_ = new TrainCaptureSampleServer(node_handle_, "train_capture_sample_server",
                        boost::bind(&CobFaceDetectionNodelet::trainCaptureSampleServerCallback, this, _1), false);
        train_capture_sample_server_->start();

        load_server_ = new LoadServer(node_handle_, "load_server", boost::bind(&CobFaceDetectionNodelet::loadServerCallback, this, _1), false);
        load_server_->start();

        save_server_ = new SaveServer(node_handle_, "save_server", boost::bind(&CobFaceDetectionNodelet::saveServerCallback, this, _1), false);
        save_server_->start();

        show_server_ = new ShowServer(node_handle_, "show_server", boost::bind(&CobFaceDetectionNodelet::showServerCallback, this, _1), false);
        show_server_->start();

        // Parameters
        std::cout << "\n--------------------------\nFace Detection Parameters:\n--------------------------\n";
        node_handle_.param("face_size_min_m", face_size_min_m_, FACE_SIZE_MIN_M);
        std::cout << "face_size_min_m = " << face_size_min_m_ << "\n";
        node_handle_.param("face_size_max_m", face_size_max_m_, FACE_SIZE_MAX_M);
        std::cout << "face_size_max_m = " << face_size_max_m_ << "\n";
        node_handle_.param("max_face_z_m", max_face_z_m_, MAX_FACE_Z_M);
        std::cout << "max_face_z_m = " << max_face_z_m_ << "\n";
        node_handle_.param("data_directory", directory_, directory_);
        std::cout << "data_directory = " << directory_ << "\n";
        node_handle_.param("fill_unassigned_depth_values", fill_unassigned_depth_values_, true);
        std::cout << "fill_unassigned_depth_values = " << fill_unassigned_depth_values_ << "\n";
        node_handle_.param("reason_about_3dface_size", reason_about_3dface_size_, true);
        std::cout << "reason_about_3dface_size = " << reason_about_3dface_size_ << "\n";

        // initializations
        init();

        //ros::spin(); not necessary with nodelets
}

unsigned long CobFaceDetectionNodelet::init()
{
        std::cout << "CobFaceDetectionNodelet::init()..." << std::endl;

        shared_image_sub_.subscribe(node_handle_, "pointcloud", 1);
        color_camera_image_sub_.subscribe(*it_, "colorimage", 1);
        //      sync_pointcloud_->connectInput(shared_image_sub_, color_camera_image_sub_);
        //      sync_pointcloud_->registerCallback(boost::bind(&CobFaceDetectionNodelet::recognizeCallback, this, _1, _2));

        sync_pointcloud_->connectInput(shared_image_sub_, color_camera_image_sub_);
        sync_pointcloud_callback_connection_ = sync_pointcloud_->registerCallback(boost::bind(&CobFaceDetectionNodelet::recognizeCallback, this, _1, _2));

        std::cout << "   inputs connected." << std::endl;

#ifdef __LINUX__
#else
        colored_pc_ = ipa_SensorFusion::CreateColoredPointCloud();
#endif

        if (people_detector_ != 0)
                delete people_detector_;
        people_detector_ = new ipa_PeopleDetector::PeopleDetector();

        std::cout << "   PeopleDetector created." << std::endl;

        filename_ = 0;

        // load data for face recognition
        loadRecognizerData();
        std::cout << "   recognizer data loaded." << std::endl;

        // use this instead if the rdata.xml file is corrupted
        // todo:
        //      loadTrainingData(false);
        //      //loadRecognizerData();
        //      //loadTrainingData(false);
        //      run_pca_ = true;
        //      PCA();
        //      saveRecognizerData();

        std::string iniFileNameAndPath = directory_ + "peopleDetectorIni.xml";

        //if (CameraSensorsControlFlow::Init(directory, "peopleDetectorIni.xml", colorCamera0, colorCamera1, rangeImagingSensor) & ipa_Utils::RET_FAILED)
        //{
        //      std::cerr << "ERROR - CameraDataViewerControlFlow::Init:" << std::endl;
        //      std::cerr << "\t ... Could not initialize 'CameraSensorsControlFlow'" << std::endl;
        //      return ipa_Utils::RET_FAILED;
        //}

        if (people_detector_->Init(directory_) & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetector::Init:" << std::endl;
                std::cerr << "\t ... Could not initialize people detector library.\n";
                return ipa_Utils::RET_FAILED;
        }
        std::cout << "   PeopleDetector initilized." << std::endl;

        if (loadParameters(iniFileNameAndPath.c_str()) & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetector::Init:" << std::endl;
                std::cerr << "\t ... Error while loading configuration file '" << std::endl;
                std::cerr << "\t ... " << iniFileNameAndPath << "'.\n";
                return ipa_Utils::RET_FAILED;
        }
        std::cout << "   ini-file loaded." << std::endl;

        run_pca_ = false;

        return ipa_Utils::RET_OK;
}

bool CobFaceDetectionNodelet::recognizeServiceServerCallback(cob_people_detection::Recognition::Request &req, cob_people_detection::Recognition::Response &res)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(2000));

        if (lock.owns_lock() == false || (occupied_by_action_ == true && recognize_server_running_ == false))
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::recognizeServerCallback:" << std::endl;
                std::cerr << "\t ... Another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                res.success = ipa_Utils::RET_FAILED;
                return false;
        }

        // read out req message
        // set up the recognition callback linkage
        if (req.running == true)
        {
                // enable recognition
                occupied_by_action_ = true;
                recognize_server_running_ = true;
                do_recognition_ = req.doRecognition;
                display_ = req.display;
                //cv::namedWindow("Face Detector");
                //cv::waitKey(1);
                //sync_pointcloud_->connectInput(shared_image_sub_, color_camera_image_sub_);
                //sync_pointcloud_callback_connection_ = sync_pointcloud_->registerCallback(boost::bind(&CobFaceDetectionNodelet::recognizeCallback, this, _1, _2, req->doRecognition, req->display));
        }
        else
        {
                // disable recognition
                occupied_by_action_ = false;
                recognize_server_running_ = false;
        }

        res.success = ipa_Utils::RET_OK;
        return true;
}

void CobFaceDetectionNodelet::recognizeServerCallback(const cob_people_detection::RecognizeGoalConstPtr& goal)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(2000));

        cob_people_detection::RecognizeResult result;
        if (lock.owns_lock() == false || (occupied_by_action_ == true && recognize_server_running_ == false))
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::recognizeServerCallback:" << std::endl;
                std::cerr << "\t ... Another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                result.success = ipa_Utils::RET_FAILED;
                recognize_server_->setSucceeded(result, "Some other action is handled at the moment.");
                return;
        }

        // read out goal message
        // set up the recognition callback linkage
        if (goal->running == true)
        {
                // enable recognition
                occupied_by_action_ = true;
                recognize_server_running_ = true;
                do_recognition_ = goal->doRecognition;
                display_ = goal->display;
                //cv::namedWindow("Face Detector");
                //cv::waitKey(1);
                //sync_pointcloud_->connectInput(shared_image_sub_, color_camera_image_sub_);
                //sync_pointcloud_callback_connection_ = sync_pointcloud_->registerCallback(boost::bind(&CobFaceDetectionNodelet::recognizeCallback, this, _1, _2, goal->doRecognition, goal->display));
        }
        else
        {
                // disable recognition
                occupied_by_action_ = false;
                recognize_server_running_ = false;
        }

        result.success = ipa_Utils::RET_OK;
        recognize_server_->setSucceeded(result);
}

void CobFaceDetectionNodelet::trainContinuousServerCallback(const cob_people_detection::TrainContinuousGoalConstPtr& goal)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(2000));

        cob_people_detection::TrainContinuousResult result;
        if (lock.owns_lock() == false || (occupied_by_action_ == true && train_continuous_server_running_ == false))
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::trainContinuousServerCallback:" << std::endl;
                std::cerr << "\t ... Another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                result.success = ipa_Utils::RET_FAILED;
                train_continuous_server_->setSucceeded(result, "Some other action is handled at the moment.");
                return;
        }

        // read out goal message
        // set up the recognition callback linkage
        if (goal->running == true)
        {
                // enable training
                capture_training_face_ = false;
                current_training_id_ = goal->trainingID;

                if (goal->appendData == true)
                        loadTrainingData(false);
                else
                {
                        filename_ = 0;
                        std::string path = directory_ + "TrainingData/";
                        try
                        {
                                boost::filesystem::remove_all(path.c_str());
                                boost::filesystem::create_directory(path.c_str());
                        } catch (const std::exception &ex)
                        {
                                std::cerr << "ERROR - PeopleDetector::trainContinuousServerCallback():" << std::endl;
                                std::cerr << "\t ... Exception catch of '" << ex.what() << "'" << std::endl;
                        }
                }
                occupied_by_action_ = true;
                train_continuous_server_running_ = true;
                //cv::namedWindow("Face Recognizer Training");
                //cv::waitKey(10);
                //sync_pointcloud_->connectInput(shared_image_sub_, color_camera_image_sub_);
                //sync_pointcloud_callback_connection_ = sync_pointcloud_->registerCallback(boost::bind(&CobFaceDetectionNodelet::trainContinuousCallback, this, _1, _2));
                std::cout << "train run...\n";

                // check whether the continuous mode was enabled or if the image capture is manually triggered
                if (goal->numberOfImagesToCapture > 0)
                {
                        lock.unlock();
                        number_training_images_captured_ = 0;
                        bool capture = true;
                        while (capture)
                        {
                                boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(10));
                                if (lock.owns_lock() && capture_training_face_ == false)
                                        capture_training_face_ = true;
                                capture = (number_training_images_captured_ < goal->numberOfImagesToCapture);
                        }

                        // turn off training mode
                        std::cout << "continuous train off...\n";

                        // disable training
                        occupied_by_action_ = false;
                        train_continuous_server_running_ = false;

                        // save images
                        saveTrainingData();

                        // do data analysis if enabled
                        if (goal->doPCA)
                        {
                                PCA();
                                saveRecognizerData();
                        }
                }
        }
        else
        {
                std::cout << "train off...\n";
                // disable training
                //sync_pointcloud_callback_connection_.disconnect();
                occupied_by_action_ = false;
                train_continuous_server_running_ = false;
                //cv::destroyWindow("Face Recognizer Training");
                //cv::waitKey(10);

                // save images
                saveTrainingData();

                // do data analysis if enabled
                if (goal->doPCA)
                {
                        PCA();
                        saveRecognizerData();
                }
        }

        result.success = ipa_Utils::RET_OK;
        train_continuous_server_->setSucceeded(result);
}

void CobFaceDetectionNodelet::trainCaptureSampleServerCallback(const cob_people_detection::TrainCaptureSampleGoalConstPtr& goal)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(10));

        cob_people_detection::TrainCaptureSampleResult result;
        if (lock.owns_lock() == false || (occupied_by_action_ == true && train_continuous_server_running_ == false) || occupied_by_action_ == false)
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::trainCaptureSampleServerCallback:" << std::endl;
                std::cerr
                                << "\t ... Either the training mode is not started or another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                result.success = ipa_Utils::RET_FAILED;
                train_capture_sample_server_->setSucceeded(result, "Some other action is handled at the moment or training mode not started.");
                return;
        }

        capture_training_face_ = true;
        result.success = ipa_Utils::RET_OK;
        train_capture_sample_server_->setSucceeded(result);
}

void CobFaceDetectionNodelet::loadServerCallback(const cob_people_detection::LoadGoalConstPtr& goal)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(2000));

        cob_people_detection::LoadResult result;
        if (lock.owns_lock() == false || occupied_by_action_ == true)
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::loadServerCallback:" << std::endl;
                std::cerr << "\t ... Another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                result.success = ipa_Utils::RET_FAILED;
                load_server_->setSucceeded(result, "Some other action is handled at the moment or training mode not started.");
                return;
        }

        // load data
        result.success = ipa_Utils::RET_OK;
        if (goal->loadMode == 0)
                result.success = loadAllData();
        else if (goal->loadMode == 1)
                result.success = loadTrainingData(goal->doPCA);
        else if (goal->loadMode == 2)
                result.success = loadRecognizerData();

        load_server_->setSucceeded(result);
}

void CobFaceDetectionNodelet::saveServerCallback(const cob_people_detection::SaveGoalConstPtr& goal)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(2000));

        cob_people_detection::SaveResult result;
        if (lock.owns_lock() == false || occupied_by_action_ == true)
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::saveServerCallback:" << std::endl;
                std::cerr << "\t ... Another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                result.success = ipa_Utils::RET_FAILED;
                save_server_->setSucceeded(result, "Some other action is handled at the moment or training mode not started.");
                return;
        }

        // save data
        result.success = ipa_Utils::RET_OK;
        if (goal->saveMode == 0)
                result.success = saveAllData();
        else if (goal->saveMode == 1)
                result.success = saveTrainingData();
        else if (goal->saveMode == 2)
                result.success = saveRecognizerData();

        save_server_->setSucceeded(result);
}

void CobFaceDetectionNodelet::showServerCallback(const cob_people_detection::ShowGoalConstPtr& goal)
{
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(2000));

        cob_people_detection::ShowResult result;
        if (lock.owns_lock() == false || occupied_by_action_ == true)
        {
                // another action is running at the moment, first the other action has to finish or to be stopped before this action can run
                std::cerr << "ERROR - PeopleDetector::showServerCallback:" << std::endl;
                std::cerr << "\t ... Another action is running at the moment. The other action has to finish or to be stopped before this action can run.\n";
                result.success = ipa_Utils::RET_FAILED;
                show_server_->setSucceeded(result, "Some other action is handled at the moment or training mode not started.");
                return;
        }

        // display
        loadTrainingData(false);
        if (goal->mode == 0)
        {
                // show average image
                cv::Mat avgImage(100, 100, CV_8UC1);
                PCA();
                showAVGImage(avgImage);

                cv::namedWindow("AVGImage");
                cv::imshow("AVGImage", avgImage);
                cv::waitKey(3000);

                cv::destroyWindow("AVGImage");
        }
        else if (goal->mode == 1)
        {
                // show eigenfaces
                cv::Mat eigenface;
                cv::namedWindow("Eigenfaces");
                PCA();
                for (int i = 0; i < (int)20/*face_images_.size()-1*/; i++)
                {
                        getEigenface(eigenface, i);
                        cv::imshow("Eigenfaces", eigenface);
                        cv::waitKey();
                }
                cv::destroyWindow("Eigenfaces");
        }

        result.success = ipa_Utils::RET_OK;
        show_server_->setSucceeded(result);
}

unsigned long CobFaceDetectionNodelet::detectFaces(cv::Mat& xyz_image, cv::Mat& color_image)
{
        cv::Mat xyz_image_8U3;
        ipa_Utils::ConvertToShowImage(xyz_image, xyz_image_8U3, 3);
        if (people_detector_->DetectFaces(color_image, xyz_image_8U3, color_faces_, range_faces_, range_face_indices_with_color_face_detection_, fill_unassigned_depth_values_)
                        & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetection::detectFaces" << std::endl;
                std::cerr << "\t ... Could not detect faces.\n";
                return ipa_Utils::RET_FAILED;
        }

        // check whether the color faces have a reasonable 3D size
        std::vector < cv::Rect > tempFaces = color_faces_; // todo: why copy to tempFaces?
        color_faces_.clear();
        // For each face...
        for (uint iface = 0; iface < tempFaces.size(); iface++)
        {
                cv::Rect& face = tempFaces[iface];

                //              one_face.box2d = faces_vec[iface];
                //              one_face.id = i; // The cascade that computed this face.

                // Get the median disparity in the middle half of the bounding box.
                int uStart = floor(0.25 * face.width);
                int uEnd = floor(0.75 * face.width) + 1;
                int vStart = floor(0.25 * face.height);
                int vEnd = floor(0.75 * face.height) + 1;
                int du = abs(uEnd - uStart);

                cv::Mat faceRegion = xyz_image(face);
                cv::Mat tmat(1, du * abs(vEnd - vStart), CV_32FC1); // vector of all depth values in the face region
                float* tmatPtr = (float*)tmat.data;
                for (int v = vStart; v < vEnd; v++)
                {
                        float* zPtr = (float*)faceRegion.row(v).data;
                        zPtr += 2 + 3 * uStart;
                        for (int u = uStart; u < uEnd; u++)
                        {
                                float depthval = *zPtr;
                                if (!isnan(depthval))
                                        *tmatPtr = depthval;
                                else
                                        *tmatPtr = 1e20;
                                tmatPtr++;
                                zPtr += 3;
                        }
                }
                //
                //
                //              int vector_position = 0;
                //              for (int v=vStart; v<vEnd; v++)
                //                for (int u=uStart; u<uEnd; u++, vector_position++)
                //                {
                //                      float depthval = xyz_image.at<cv::Vec3f>(v,u).z;
                //                      if (!isnan(depthval)) tmat.at<float>(0,vector_position) = depthval;
                //                      else tmat.at<float>(0,vector_position) = 1e20;
                //                }

                cv::Mat tmat_sorted;
                cv::sort(tmat, tmat_sorted, CV_SORT_EVERY_COLUMN + CV_SORT_DESCENDING);
                double avg_depth = tmat_sorted.at<float>(floor(cv::countNonZero(tmat_sorted >= 0.0) * 0.5)); // Get the middle valid disparity (-1 disparities are invalid)

                // Fill in the rest of the face data structure.
                //              one_face.center2d = cv::Point2d(one_face.box2d.x+one_face.box2d.width*0.5,
                //                                                                              one_face.box2d.y+one_face.box2d.height*0.5);
                //              one_face.radius2d = one_face.box2d.width*0.5;

                // If the median disparity was valid and the face is a reasonable size, the face status is "good".
                // If the median disparity was valid but the face isn't a reasonable size, the face status is "bad".
                // Otherwise, the face status is "unknown".
                // Only bad faces are removed
                if (avg_depth > 0)
                {
                        double radiusX, radiusY, radius3d = 1e20;
                        cv::Vec3f a, b;
                        // vertical line regularly lies completely on the head whereas this does not hold very often for the horizontal line crossing the bounding box of the face
                        // rectangle in the middle
                        a = xyz_image.at<cv::Vec3f>((int)(face.y + face.height * 0.25), (int)(face.x + 0.5 * face.width));
                        b = xyz_image.at<cv::Vec3f>((int)(face.y + face.height * 0.75), (int)(face.x + 0.5 * face.width));
                        if (display_)
                                std::cout << "a: " << a.val[0] << " " << a.val[1] << " " << a.val[2] << "   b: " << " " << b.val[0] << " " << b.val[1] << " " << b.val[2] << "\n";
                        if (isnan(a.val[0]) || isnan(b.val[0]))
                                radiusY = 0.0;
                        else
                                radiusY = cv::norm(b - a);
                        radius3d = radiusY;

                        // for radius estimation with the horizontal line through the face rectangle use points which typically still lie on the face and not in the background
                        a = xyz_image.at<cv::Vec3f>((int)(face.y + face.height * 0.5), (int)(face.x + face.width * 0.25));
                        b = xyz_image.at<cv::Vec3f>((int)(face.y + face.height * 0.5), (int)(face.x + face.width * 0.75));
                        if (display_)
                                std::cout << "a: " << a.val[0] << " " << a.val[1] << " " << a.val[2] << "   b: " << " " << b.val[0] << " " << b.val[1] << " " << b.val[2] << "\n";
                        if (isnan(a.val[0]) || isnan(b.val[0]))
                                radiusX = 0.0;
                        else
                        {
                                radiusX = cv::norm(b - a);
                                if (radiusY != 0.0)
                                        radius3d = (radiusX + radiusY) * 0.5;
                                else
                                        radius3d = radiusX;
                        }

                        //                      cv::Point pup(face.x+0.5*face.width, face.y+face.height*0.25);
                        //                      cv::Point plo(face.x+0.5*face.width, face.y+face.height*0.75);
                        //                      cv::Point ple(face.x+face.width*0.25, face.y+face.height*0.5);
                        //                      cv::Point pri(face.x+face.width*0.75, face.y+face.height*0.5);
                        //                      cv::line(xyz_image_8U3, pup, plo, CV_RGB(255, 255, 255), 2);
                        //                      cv::line(xyz_image_8U3, ple, pri, CV_RGB(255, 255, 255), 2);

                        if (display_)
                        {
                                std::cout << "radiusX: " << radiusX << "  radiusY: " << radiusY << "\n";
                                std::cout << "avg_depth: " << avg_depth << " > max_face_z_m_: " << max_face_z_m_ << " ?  2*radius3d: " << 2.0 * radius3d << " < face_size_min_m_: "
                                                << face_size_min_m_ << " ?  2radius3d: " << 2.0 * radius3d << " > face_size_max_m_:" << face_size_max_m_ << "?\n";
                        }
                        if (reason_about_3dface_size_ == true && radius3d > 0.0 && (avg_depth > max_face_z_m_ || 2.0 * radius3d < face_size_min_m_ || 2.0 * radius3d > face_size_max_m_))
                                continue; // face does not match normal human appearance
                }
                color_faces_.push_back(face);
        }
        //                      imshow("xyz image", xyz_image_8U3);
        //                      cv::waitKey(10);

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::recognizeFace(cv::Mat& color_image, std::vector<int>& index)
{
        if (people_detector_->RecognizeFace(color_image, color_faces_, &n_eigens_, eigen_vectors_, avg_image_, face_class_avg_projections_, index, &threshold_, &threshold_fs_,
                        eigen_val_mat_, /*&person_classifier_*/0) & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetector::recognizeFace:" << std::endl;
                std::cerr << "\t ... Error while recognizing faces.\n";
                return ipa_Utils::RET_FAILED;
        }

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::addFace(cv::Mat& image, std::string id)
{
        // addFace should only be called if there is exactly one face found --> so we access it with color_faces_[0]
        if (people_detector_->AddFace(image, color_faces_[0], id, face_images_, id_) & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetectorControlFlow::AddFace:" << std::endl;
                std::cerr << "\t ... Could not save face.\n";
                return ipa_Utils::RET_FAILED;
        }
        run_pca_ = true;

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::PCA()
{
        // only run PCA if new data has arrived after last computation
        if (!run_pca_)
        {
                //std::cout << "INFO - PeopleDetector::PCA:" << std::endl;
                //std::cout << "\t ... PCA algorithm skipped.\n";
                return ipa_Utils::RET_OK;
        }

        if (face_images_.size() < 2)
        {
                std::cout << "WARNING - PeopleDetector::ConsoleGUI:" << std::endl;
                std::cout << "\t ... Less than two images are trained.\n";
                return ipa_Utils::RET_OK;
        }

        // Delete memory
        eigen_vectors_.clear();

        // mirror images
        //todo:
        //      std::vector<cv::Mat> face_images_doubled;
        //      std::vector<std::string> id_doubled;
        //      for (int i=0; i<(int)face_images_.size(); i++)
        //      {
        //              face_images_doubled.push_back(face_images_[i]);
        //              id_doubled.push_back(id_[i]);
        //              cv::Mat temp;
        //              cv::flip(face_images_[i], temp, 1);
        //              face_images_doubled.push_back(temp);
        //              id_doubled.push_back(id_[i]);
        //      }

        // Do PCA
        if (people_detector_->PCA(&n_eigens_, eigen_vectors_, eigen_val_mat_, avg_image_, face_images_/*doubled*/, projected_train_face_mat_) & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetectorControlFlow::PCA:" << std::endl;
                std::cerr << "\t ... Error while PCA.\n";
                return ipa_Utils::RET_FAILED;
        }

        // Calculate FaceClasses
        std::cout << "Debug: n_eigens: " << n_eigens_ << " id: " << id_.size() << "\n";
        if (people_detector_->CalculateFaceClasses(projected_train_face_mat_, id_/*doubled*/, &n_eigens_, face_class_avg_projections_, id_unique_, /*&person_classifier_*/0)
                        & ipa_Utils::RET_FAILED)
        {
                std::cerr << "ERROR - PeopleDetectorControlFlow::PCA:" << std::endl;
                std::cerr << "\t ... Error while calculating FaceClasses.\n";
                return ipa_Utils::RET_FAILED;
        }

        run_pca_ = false;

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::saveAllData()
{
        //      try
        //      {
        //              fs::remove_all(path.c_str());
        //              fs::create_directory(path.c_str());
        //      }
        //      catch (const std::exception &ex)
        //      {
        //              std::cerr << "ERROR - PeopleDetector::SaveTrainingData():" << std::endl;
        //              std::cerr << "\t ... Exception catch of '" << ex.what() << "'" << std::endl;
        //      }

        if (saveTrainingData() != ipa_Utils::RET_OK)
                return ipa_Utils::RET_FAILED;
        if (saveRecognizerData() != ipa_Utils::RET_OK)
                return ipa_Utils::RET_FAILED;

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::saveTrainingData()
{
        std::string path = directory_ + "TrainingData/";
        std::string filename = "tdata.xml";
        std::string img_ext = ".bmp";

        std::ostringstream complete;
        complete << path << filename;

        cv::FileStorage fileStorage(complete.str().c_str(), cv::FileStorage::WRITE);
        if (!fileStorage.isOpened())
        {
                std::cout << "WARNING - PeopleDetector::SaveTrainingData:" << std::endl;
                std::cout << "\t ... Can't save training data.\n";
                return ipa_Utils::RET_OK;
        }

        // Ids
        fileStorage << "id_num" << (int)id_.size();
        for (int i = 0; i < (int)id_.size(); i++)
        {
                std::ostringstream tag;
                tag << "id_" << i;
                fileStorage << tag.str().c_str() << id_[i].c_str();
        }

        // Face images
        fileStorage << "faces_num" << (int)face_images_.size();
        for (int i = 0; i < (int)face_images_.size(); i++)
        {
                std::ostringstream img, shortname;
                img << path << i << img_ext;
                shortname << "TrainingData/" << i << img_ext;
                std::ostringstream tag;
                tag << "img_" << i;
                fileStorage << tag.str().c_str() << shortname.str().c_str();
                cv::imwrite(img.str().c_str(), face_images_[i]);
        }

        fileStorage.release();

        std::cout << "INFO - PeopleDetector::SaveTrainingData:" << std::endl;
        std::cout << "\t ... " << face_images_.size() << " images saved.\n";
        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::saveRecognizerData()
{
        std::string path = directory_ + "TrainingData/";
        std::string filename = "rdata.xml";

        std::ostringstream complete;
        complete << path << filename;

        cv::FileStorage fileStorage(complete.str().c_str(), cv::FileStorage::WRITE);
        if (!fileStorage.isOpened())
        {
                std::cout << "WARNING - PeopleDetector::saveRecognizerData:" << std::endl;
                std::cout << "\t ... Can't save training data.\n";
                return ipa_Utils::RET_OK;
        }

        // Number eigenvalues/eigenvectors
        fileStorage << "eigens_num" << n_eigens_;

        // Eigenvectors
        for (int i = 0; i < n_eigens_; i++)
        {
                std::ostringstream tag;
                tag << "ev_" << i;
                fileStorage << tag.str().c_str() << eigen_vectors_[i];
        }

        // Eigenvalue matrix
        fileStorage << "eigen_val_mat" << eigen_val_mat_;

        // Average image
        fileStorage << "avg_image" << avg_image_;

        // Projection coefficients of the training faces
        fileStorage << "projected_train_face_mat" << projected_train_face_mat_;

        // Unique Ids (id_unique_[i] stores the corresponding id to the average face coordinates in the face subspace in face_class_avg_projections_.row(i))
        fileStorage << "id_unique_num" << (int)id_unique_.size();
        for (int i = 0; i < (int)id_unique_.size(); i++)
        {
                std::ostringstream tag;
                tag << "id_unique_" << i;
                fileStorage << tag.str().c_str() << id_unique_[i].c_str();
        }

        // The average factors of the eigenvector decomposition from each face class
        fileStorage << "face_class_avg_projections" << face_class_avg_projections_;

        fileStorage.release();

        // save classifier
        std::string classifierFile = path + "svm.dat";
        //person_classifier_.save(classifierFile.c_str());

        std::cout << "INFO - PeopleDetector::saveRecognizerData:" << std::endl;
        std::cout << "\t ... recognizer data saved.\n";
        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::loadAllData()
{
        if (loadTrainingData(true) != ipa_Utils::RET_OK)
                return ipa_Utils::RET_FAILED;
        if (loadRecognizerData() != ipa_Utils::RET_OK)
                return ipa_Utils::RET_FAILED;

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::loadTrainingData(bool runPCA)
{
        std::string path = directory_ + "TrainingData/";
        std::string filename = "tdata.xml";

        std::ostringstream complete;
        complete << path << filename;

        if (fs::is_directory(path.c_str()))
        {
                cv::FileStorage fileStorage(complete.str().c_str(), cv::FileStorage::READ);
                if (!fileStorage.isOpened())
                {
                        std::cout << "WARNING - PeopleDetector::loadTrainingData:" << std::endl;
                        std::cout << "\t ... Cant open " << complete.str() << ".\n";
                        return ipa_Utils::RET_OK;
                }

                // Ids
                id_.clear();
                int id_num = (int)fileStorage["id_num"];
                for (int i = 0; i < id_num; i++)
                {
                        std::ostringstream tag;
                        tag << "id_" << i;
                        id_.push_back((std::string)fileStorage[tag.str().c_str()]);
                }

                // Images
                face_images_.clear();
                int faces_num = (int)fileStorage["faces_num"];
                for (int i = 0; i < faces_num; i++)
                {
                        std::ostringstream tag;
                        tag << "img_" << i;
                        std::string path = directory_ + (std::string)fileStorage[tag.str().c_str()];
                        cv::Mat temp = cv::imread(path.c_str(), 0);
                        face_images_.push_back(temp);
                }

                // set next free filename
                filename_ = faces_num;

                fileStorage.release();

                // Run PCA - now or later
                run_pca_ = true;
                if (runPCA)
                        PCA();

                std::cout << "INFO - PeopleDetector::loadTrainingData:" << std::endl;
                std::cout << "\t ... " << faces_num << " images loaded.\n";
        }
        else
        {
                std::cerr << "ERROR - PeopleDetector::loadTrainingData:" << std::endl;
                std::cerr << "\t .... Path '" << path << "' is not a directory." << std::endl;
                return ipa_Utils::RET_FAILED;
        }

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::loadRecognizerData()
{
        std::string path = directory_ + "TrainingData/";
        std::string filename = "rdata.xml";

        std::ostringstream complete;
        complete << path << filename;

        if (fs::is_directory(path.c_str()))
        {
                cv::FileStorage fileStorage(complete.str().c_str(), cv::FileStorage::READ);
                if (!fileStorage.isOpened())
                {
                        std::cout << "WARNING - PeopleDetector::loadRecognizerData:" << std::endl;
                        std::cout << "\t ... Cant open " << complete.str() << ".\n";
                        return ipa_Utils::RET_OK;
                }
                std::cout << "   loading recognizer data..." << std::endl;

                // Number eigenvalues/eigenvectors
                n_eigens_ = (int)fileStorage["eigens_num"];

                // Eigenvectors
                eigen_vectors_.clear();
                eigen_vectors_.resize(n_eigens_, cv::Mat());
                for (int i = 0; i < n_eigens_; i++)
                {
                        std::ostringstream tag;
                        tag << "ev_" << i;
                        fileStorage[tag.str().c_str()] >> eigen_vectors_[i];
                }

                // Eigenvalue matrix
                eigen_val_mat_ = cv::Mat();
                fileStorage["eigen_val_mat"] >> eigen_val_mat_;

                // Average image
                avg_image_ = cv::Mat();
                fileStorage["avg_image"] >> avg_image_;

                // Projections of the training faces
                projected_train_face_mat_ = cv::Mat();
                fileStorage["projected_train_face_mat"] >> projected_train_face_mat_;

                // Unique Ids (id_unique_[i] stores the corresponding id to the average face coordinates in the face subspace in face_class_avg_projections_.row(i))
                id_unique_.clear();
                int idUniqueNum = (int)fileStorage["id_unique_num"];
                for (int i = 0; i < idUniqueNum; i++)
                {
                        std::ostringstream tag;
                        tag << "id_unique_" << i;
                        id_unique_.push_back((std::string)fileStorage[tag.str().c_str()]);
                }

                // The average factors of the eigenvector decomposition from each face class
                face_class_avg_projections_ = cv::Mat();
                fileStorage["face_class_avg_projections"] >> face_class_avg_projections_;

                fileStorage.release();

                // load classifier
                std::string classifierFile = path + "svm.dat";
                //person_classifier_.load(classifierFile.c_str());

                // do not run PCA
                run_pca_ = false;

                std::cout << "INFO - PeopleDetector::loadRecognizerData:" << std::endl;
                std::cout << "\t ... recognizer data loaded.\n";
        }
        else
        {
                std::cerr << "ERROR - PeopleDetector::loadRecognizerData():" << std::endl;
                std::cerr << "\t .... Path '" << path << "' is not a directory." << std::endl;
                return ipa_Utils::RET_FAILED;
        }

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::getEigenface(cv::Mat& eigenface, int index)
{
        cv::normalize(eigen_vectors_[index], eigenface, 0, 255, cv::NORM_MINMAX, CV_8UC1);

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::showAVGImage(cv::Mat& avgImage)
{
        if (!run_pca_ && face_images_.size() < 2)
        {
                std::cerr << "PeopleDetector::showAvgImage()" << std::endl;
                std::cerr << "No AVG image calculated" << std::endl;
                return 0;
        }

        cv::normalize(avg_image_, avgImage, 0, 255, cv::NORM_MINMAX, CV_8UC1);

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::saveRangeTrainImages(cv::Mat& xyz_image)
{
        std::string path = directory_ + "haarcascades/";
        std::string img_ext = ".jpg";
        cv::Mat xyzImage_8U3(xyz_image.size(), CV_8UC3); //IplImage* xyzImage_8U3 = cvCreateImage(cvGetSize(pc->GetXYZImage()), 8, 3);
        ipa_Utils::ConvertToShowImage(xyz_image, xyzImage_8U3, 3);

        for (int i = 0; i < (int)color_faces_.size(); i++)
        {
                cv::Mat xyzImage_8U3_resized(100, 100, CV_8UC3); //cvCreateImage(cvSize(100,100), 8, 3); 8=IPL_DEPTH_8U

                double scale = 1.6;
                cv::Rect rangeFace;
                rangeFace.height = (int)(color_faces_[i].height * scale);
                rangeFace.width = (int)(color_faces_[i].width * scale);
                rangeFace.x = color_faces_[i].x - ((rangeFace.width - color_faces_[i].width) / 2);
                rangeFace.y = color_faces_[i].y - ((rangeFace.height - color_faces_[i].height) / 2) - 10;

                cv::Mat xyzImage_8U3_roi = xyzImage_8U3(rangeFace);
                //cvSetImageROI(xyzImage_8U3, rangeFace);
                cv::resize(xyzImage_8U3_roi, xyzImage_8U3_resized, xyzImage_8U3_resized.size());

                std::ostringstream file;
                file << path << filename_ << img_ext;

                cv::imwrite(file.str().c_str(), xyzImage_8U3_resized);//cvSaveImage(file.str().c_str(), xyzImage_8U3_resized);
                filename_++;
        }

        return ipa_Utils::RET_OK;
}

//unsigned long CobFaceDetectionNodelet::getMeasurement(const sensor_msgs::PointCloud2::ConstPtr& shared_image_msg, const sensor_msgs::Image::ConstPtr& color_image_msg)
//{
//      cv::Mat color_image_8U3(shared_image_msg->height, shared_image_msg->width, CV_8UC3);
//      cv::Mat xyz_image_32F3(shared_image_msg->height, shared_image_msg->width, CV_32FC3);
//      float* f_ptr = 0;
//      const uint8_t* data_ptr = 0;
//      unsigned char* uc_ptr = 0;
//      unsigned int xyz_offset = shared_image_msg->fields[0].offset;
//      unsigned int rgb_offset = shared_image_msg->fields[3].offset;
//      size_t b_offset = 2*sizeof(unsigned char);
//      size_t g_offset = sizeof(unsigned char);
//      size_t r_offset = 0;
//      unsigned int col_times_3 = 0;
//      for (unsigned int row = 0; row < shared_image_msg->height; row++)
//      {
//              uc_ptr = color_image_8U3.ptr<unsigned char>(row);
//              f_ptr = xyz_image_32F3.ptr<float>(row);
//
//              data_ptr = &shared_image_msg->data[row * shared_image_msg->width * shared_image_msg->point_step];
//
//              for (unsigned int col = 0; col < shared_image_msg->width; col++)
//              {
//                      col_times_3 = 3*col;
//                      // Reorder incoming image channels
//                      memcpy(&uc_ptr[col_times_3], &data_ptr[col * shared_image_msg->point_step + rgb_offset + b_offset], sizeof(unsigned char));
//                      memcpy(&uc_ptr[col_times_3 + 1], &data_ptr[col * shared_image_msg->point_step + rgb_offset + g_offset], sizeof(unsigned char));
//                      memcpy(&uc_ptr[col_times_3 + 2], &data_ptr[col * shared_image_msg->point_step + rgb_offset + r_offset], sizeof(unsigned char));
//
//                      memcpy(&f_ptr[col_times_3], &data_ptr[col * shared_image_msg->point_step + xyz_offset], 3*sizeof(float));
//              }
//      }
//
//#ifdef __LINUX__
//      color_image_ = color_image_8U3;
//      range_image_ = xyz_image_32F3;
//#else
//      // Images are cloned within setter functions
//      colored_pc_->SetColorImage(color_image_8U3);
//      colored_pc_->SetXYZImage(xyz_image_32F3);
//#endif
//
//
//      return ipa_Utils::RET_OK;
//}


unsigned long CobFaceDetectionNodelet::convertColorImageMessageToMat(const sensor_msgs::Image::ConstPtr& color_image_msg, cv_bridge::CvImageConstPtr& color_image_ptr,
                cv::Mat& color_image)
{
        try
        {
                color_image_ptr = cv_bridge::toCvShare(color_image_msg, sensor_msgs::image_encodings::BGR8);
        } catch (cv_bridge::Exception& e)
        {
                ROS_ERROR("PeopleDetection: cv_bridge exception: %s", e.what());
                return ipa_Utils::RET_FAILED;
        }
        color_image = color_image_ptr->image;

        return ipa_Utils::RET_OK;
}

unsigned long CobFaceDetectionNodelet::convertPclMessageToMat(const sensor_msgs::PointCloud2::ConstPtr& shared_image_msg, cv::Mat& depth_image)
{
        pcl::PointCloud < pcl::PointXYZ > depth_cloud; // point cloud
        pcl::fromROSMsg(*shared_image_msg, depth_cloud);
        depth_image.create(depth_cloud.height, depth_cloud.width, CV_32FC3);
        uchar* depth_image_ptr = (uchar*)depth_image.data;
        for (int v = 0; v < (int)depth_cloud.height; v++)
        {
                int baseIndex = depth_image.step * v;
                for (int u = 0; u < (int)depth_cloud.width; u++)
                {
                        int index = baseIndex + 3 * u * sizeof(float);
                        float* data_ptr = (float*)(depth_image_ptr + index);
                        pcl::PointXYZ point_xyz = depth_cloud(u, v);
                        data_ptr[0] = point_xyz.x;
                        data_ptr[1] = point_xyz.y;
                        data_ptr[2] = (isnan(point_xyz.z)) ? 0.f : point_xyz.z;
                        //if (u%100 == 0) std::cout << "u" << u << " v" << v << " z" << data_ptr[2] << "\n";
                }
        }
        return ipa_Utils::RET_OK;
}

inline std::string CobFaceDetectionNodelet::getLabel(int index)
{
        switch (index)
        {
        case -1:
                // Distance to face class is too high
                return "Unknown";
                break;
        case -2:
                // Distance to face space is too high
                return "No face";
                break;
        default:
                // Face classified
                return id_unique_[index];
        }
}

inline int abs(int x)
{
        return ((x < 0) ? -x : x);
}

void CobFaceDetectionNodelet::recognizeCallback(const sensor_msgs::PointCloud2::ConstPtr& shared_image_msg, const sensor_msgs::Image::ConstPtr& color_image_msg)
{
        // check if this is a training call
        if (train_continuous_server_running_ == true)
        {
                trainContinuousCallback(shared_image_msg, color_image_msg);
                return;
        }

        // check for disabled recognition
        if (recognize_server_running_ == false)
                return;

        // convert input to cv::Mat images
        // color
        cv_bridge::CvImageConstPtr color_image_ptr;
        cv::Mat color_image;
        convertColorImageMessageToMat(color_image_msg, color_image_ptr, color_image);//color_image_ptr->image;

        // point cloud
        cv::Mat depth_image;
        convertPclMessageToMat(shared_image_msg, depth_image);

#ifdef __LINUX__
        color_image_ = color_image;
        range_image_ = depth_image;
#else
        // convert point cloud and color image to colored point cloud
        colored_pc_->SetColorImage(color_image);
        colored_pc_->SetXYZImage(depth_image);
#endif
        //getMeasurement(shared_image_msg, color_image_msg);

        // todo: should not be necessary here (e.g. set a trained flag)
        PCA();

        if (eigen_vectors_.size() < 1)
        {
                std::cout << "WARNING - PeopleDetector::ConsoleGUI:" << std::endl;
                std::cout << "\t ... Less than two images are trained.\n";
                return;
        }

        //DWORD start = timeGetTime();
        cv::Mat colorImage_8U3;
#ifdef __LINUX__
        detectFaces(range_image_, color_image_);
        colorImage_8U3 = color_image_.clone();
#else
        detectFaces(colored_pc_->GetXYZImage(), colored_pc_->GetColorImage());
        colored_pc_->GetColorImage().copyTo(colorImage_8U3);
#endif

        std::vector<int> index;
        if (do_recognition_ == true)
        {
                recognizeFace(color_image, index);
                //std::cout << "INFO - PeopleDetector::Recognize:" << std::endl;
        }
        else
                for (int i = 0; i < (int)color_faces_.size(); i++)
                        index.push_back(-1);
        //std::cout << "\t ... Recognize Time: " << (timeGetTime() - start) << std::endl;

        // publish face positions
        std::stringstream ss;
        ss << depth_image.rows << " " << depth_image.cols;
        cob_perception_msgs::DetectionArray facePositionMsg;
        // image dimensions
        facePositionMsg.header.frame_id = ss.str();
        // time stamp
        facePositionMsg.header.stamp = ros::Time::now(); //color_image_msg->header.stamp;  //
        //facePositionMsg.detections.reserve(color_faces_.size());
        // add all range faces that do not contain a face detection in the color image
        for (int i = 0; i < (int)range_faces_.size(); i++)
        {
                // if no color face was detected in that region, add it to the published list
                if (range_face_indices_with_color_face_detection_.find(i) == range_face_indices_with_color_face_detection_.end())
                {
                        cv::Rect face = range_faces_[i];

                        // 2D image coordinates
                        cob_perception_msgs::Detection det;
                        det.mask.roi.x = face.x;
                        det.mask.roi.y = face.y;
                        det.mask.roi.width = face.width;
                        det.mask.roi.height = face.height;
                        float center2Dx = face.x + face.width * 0.5f;
                        float center2Dy = face.y + face.height * 0.5f;

                        // 3D world coordinates (and verify that the read pixel contained valid coordinates, otherwise search for valid pixel in neighborhood)
                        geometry_msgs::Point* point = &(det.pose.pose.position);
                        cv::Point3f p;
                        bool validCoordinates = false;
                        for (int d = 0; (d < 6 && !validCoordinates); d++)
                        {
                                for (int v = -d; (v <= d && !validCoordinates); v++)
                                {
                                        for (int u = -d; (u <= d && !validCoordinates); u++)
                                        {
                                                if (abs(v) != d && abs(u) != d)
                                                        continue;

                                                p = depth_image.at<cv::Point3f>(center2Dy + v, center2Dx + u);
                                                point->x = p.x;
                                                point->y = p.y;
                                                point->z = p.z;

                                                if (!isnan(point->x) && !isnan(point->y) && point->z != 0.f)
                                                        validCoordinates = true;
                                        }
                                }
                        }
                        if (validCoordinates == false)
                                continue;

                        // person ID
                        det.label = "UnknownRange";

                        // origin of detection
                        det.detector = "range";

                        facePositionMsg.detections.push_back(det);
                }
        }

        // add all color face detections
        for (int i = 0; i < (int)color_faces_.size(); i++)
        {
                cv::Rect face = color_faces_[i];

                // 2D image coordinates
                cob_perception_msgs::Detection det;
                det.mask.roi.x = face.x;
                det.mask.roi.y = face.y;
                det.mask.roi.width = face.width;
                det.mask.roi.height = face.height;
                float center2Dx = face.x + face.width * 0.5f;
                float center2Dy = face.y + face.height * 0.5f;

                // 3D world coordinates
                //              cv::Point3f p = depth_image.at<cv::Point3f>(center2Dy, center2Dx);
                //              geometry_msgs::Point* point = &(det.pose.pose.position);
                //              point->x = p.x; point->y=p.y; point->z=p.z;

                // 3D world coordinates (and verify that the read pixel contained valid coordinates, otherwise search for valid pixel in neighborhood)
                geometry_msgs::Point* point = &(det.pose.pose.position);
                cv::Point3f p;
                bool validCoordinates = false;
                for (int d = 0; (d < 6 && !validCoordinates); d++)
                {
                        for (int v = -d; (v <= d && !validCoordinates); v++)
                        {
                                for (int u = -d; (u <= d && !validCoordinates); u++)
                                {
                                        if (abs(v) != d && abs(u) != d)
                                                continue;

                                        p = depth_image.at<cv::Point3f>(center2Dy + v, center2Dx + u);
                                        point->x = p.x;
                                        point->y = p.y;
                                        point->z = p.z;

                                        if (!isnan(point->x) && !isnan(point->y) && point->z != 0.f)
                                                validCoordinates = true;
                                }
                        }
                }
                if (validCoordinates == false)
                        continue;

                // person ID
                det.label = getLabel(index[i]);

                // origin of detection
                det.detector = "color";

                facePositionMsg.detections.push_back(det);
        }
        face_position_publisher_.publish(facePositionMsg);

        //      std_msgs::Float32MultiArrayPtr facePositionMsg(new std_msgs::Float32MultiArray);
        //      const int dataBlockSize = 8;
        //      std::vector<float> data(dataBlockSize*color_faces_.size());
        //      for(int i=0; i<(int)color_faces_.size(); i++)
        //      {
        //              cv::Rect face = color_faces_[i];
        //              int baseIndex = i*dataBlockSize;
        //
        //              // 2D image coordinates
        //              data[baseIndex]=face.x; data[baseIndex+1]=face.y; data[baseIndex+2]=face.width; data[baseIndex+3]=face.height;
        //              float center2Dx = face.x + face.width*0.5f;
        //              float center2Dy = face.y + face.height*0.5f;
        //
        //              // 3D world coordinates
        //              cv::Point3f p = depth_image.at<cv::Point3f>(center2Dy, center2Dx);
        //              data[baseIndex+4]=p.x; data[baseIndex+5]=p.y; data[baseIndex+6]=p.z;
        //
        //              // person ID
        //              data[baseIndex+7]=index[i];
        //      }
        //      facePositionMsg->data = data;
        //      std::vector<std_msgs::MultiArrayDimension> dim(2, std_msgs::MultiArrayDimension());
        //      dim[0].label = "faces";       dim[0].size = color_faces_.size();        dim[0].stride = data.size();
        //      dim[1].label = "coordinates"; dim[1].size = dataBlockSize;          dim[1].stride = dataBlockSize;
        //      std_msgs::MultiArrayLayout layout;
        //      layout.dim = dim;
        //      layout.data_offset = 0;
        //      facePositionMsg->layout = layout;
        //      face_position_publisher_->publish(facePositionMsg);

        // display results
        if (display_ == true)
        {
                for (int i = 0; i < (int)range_faces_.size(); i++)
                {
                        cv::Rect face = range_faces_[i];
                        cv::rectangle(colorImage_8U3, cv::Point(face.x, face.y), cv::Point(face.x + face.width, face.y + face.height), CV_RGB(0, 0, 255), 2, 8, 0);
                }

                for (int i = 0; i < (int)color_faces_.size(); i++)
                {
                        cv::Rect face = color_faces_[i];

                        cv::rectangle(colorImage_8U3, cv::Point(face.x, face.y), cv::Point(face.x + face.width, face.y + face.height), CV_RGB(0, 255, 0), 2, 8, 0);

                        std::stringstream tmp;
                        switch (index[i])
                        {
                        case -1:
                                // Distance to face class is too high
                                tmp << "Unknown";
                                cv::putText(colorImage_8U3, tmp.str().c_str(), cv::Point(face.x, face.y + face.height + 25), cv::FONT_HERSHEY_PLAIN, 2, CV_RGB( 255, 0, 0 ), 2);
                                break;
                        case -2:
                                // Distance to face space is too high
                                tmp << "No face";
                                cv::putText(colorImage_8U3, tmp.str().c_str(), cv::Point(face.x, face.y + face.height + 25), cv::FONT_HERSHEY_PLAIN, 2, CV_RGB( 255, 0, 0 ), 2);
                                break;
                        default:
                                // Face classified
                                tmp << id_unique_[index[i]].c_str();
                                cv::putText(colorImage_8U3, tmp.str().c_str(), cv::Point(face.x, face.y + face.height + 25), cv::FONT_HERSHEY_PLAIN, 2, CV_RGB( 0, 255, 0 ), 2);
                        }
                }
                //cv::imshow("Face Detector", colorImage_8U3);
                //cv::waitKey(10);
                // publish topic
                cv_bridge::CvImage cv_ptr;
                cv_ptr.image = colorImage_8U3;
                cv_ptr.encoding = "bgr8";
                face_detection_image_pub_.publish(cv_ptr.toImageMsg());
        }
}

void CobFaceDetectionNodelet::trainContinuousCallback(const sensor_msgs::PointCloud2::ConstPtr& shared_image_msg, const sensor_msgs::Image::ConstPtr& color_image_msg)
{
        if (train_continuous_server_running_ == false)
                return;

        // convert input to cv::Mat images
        // color
        cv_bridge::CvImageConstPtr color_image_ptr;
        cv::Mat color_image;
        convertColorImageMessageToMat(color_image_msg, color_image_ptr, color_image);//color_image_ptr->image;

        // point cloud
        cv::Mat depth_image;
        convertPclMessageToMat(shared_image_msg, depth_image);

#ifdef __LINUX__
        color_image_ = color_image;
        range_image_ = depth_image;
#else
        // convert point cloud and color image to colored point cloud
        colored_pc_->SetColorImage(color_image);
        colored_pc_->SetXYZImage(depth_image);
#endif
        //getMeasurement(shared_image_msg, color_image_msg);

        cv::Mat colorImage_8U3;
#ifdef __LINUX__
        detectFaces(range_image_, color_image_);
        colorImage_8U3 = color_image_.clone();
#else
        detectFaces(colored_pc_->GetXYZImage(), colored_pc_->GetColorImage());
        colored_pc_->GetColorImage().copyTo(colorImage_8U3);
#endif

        for (int i = 0; i < (int)color_faces_.size(); i++)
        {
                cv::Rect face = color_faces_[i];
                cv::rectangle(colorImage_8U3, cv::Point(face.x, face.y), cv::Point(face.x + face.width, face.y + face.height), CV_RGB( 0, 255, 0 ), 2, 8, 0);
        }

        //cv::imshow("Face Recognizer Training", colorImage_8U3);
        //cv::waitKey(10);
        // publish topic
        cv_bridge::CvImage cv_ptr;
        cv_ptr.image = colorImage_8U3;
        cv_ptr.encoding = "bgr8";
        face_detection_image_pub_.publish(cv_ptr.toImageMsg());

        // capture image if triggered by an action
        // secure this section with a mutex
        boost::timed_mutex::scoped_timed_lock lock(action_mutex_, boost::posix_time::milliseconds(10));
        if (lock.owns_lock() && capture_training_face_ == true)
        {
                capture_training_face_ = false;

                // Check if there is more than one face detected
                if (color_faces_.size() > 1)
                {
                        std::cout << "WARNING - CobFaceDetectionNodelet::trainContinuousCallback:" << std::endl;
                        std::cout << "\t ... More than one faces are detected in image. Please try again." << std::endl;
                        return;
                }

                // Check if there is less than one face detected
                if (color_faces_.size() < 1)
                {
                        std::cout << "WARNING - CobFaceDetectionNodelet::trainContinuousCallback:" << std::endl;
                        std::cout << "\t ... Less than one faces are detected in image. Please try again." << std::endl;
                        return;
                }

#ifdef __LINUX__
                addFace(color_image_, current_training_id_);
#else
                addFace(colored_pc_->GetColorImage(), current_training_id_);
#endif
                number_training_images_captured_++;
                std::cout << "INFO - CuiPeopleDetector::ConsoleGUI:" << std::endl;
                std::cout << "\t ... Face captured (" << face_images_.size() << ")." << std::endl;
        }
}

unsigned long CobFaceDetectionNodelet::loadParameters(const char* iniFileName)
{
        TiXmlDocument* p_configXmlDocument = new TiXmlDocument(iniFileName);
        if (!p_configXmlDocument->LoadFile())
        {
                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                std::cerr << "\t ... Error while loading xml configuration file" << std::endl;
                std::cerr << "\t ... (Check filename and syntax of the file):" << std::endl;
                std::cerr << "\t ... '" << iniFileName << "'" << std::endl;
                return ipa_Utils::RET_FAILED;
        }
        std::cout << "INFO - PeopleDetector::LoadParameters:" << std::endl;
        std::cout << "\t ... Parsing xml configuration file:" << std::endl;
        std::cout << "\t ... '" << iniFileName << "'" << std::endl;

        if (p_configXmlDocument)
        {

                //************************************************************************************
                //      BEGIN PeopleDetector
                //************************************************************************************
                // Tag element "PeopleDetector" of Xml Inifile
                TiXmlElement *p_xmlElement_Root = NULL;
                p_xmlElement_Root = p_configXmlDocument->FirstChildElement("PeopleDetector");

                if (p_xmlElement_Root)
                {

                        //************************************************************************************
                        //      BEGIN PeopleDetector->adaBoost
                        //************************************************************************************
                        // Tag element "adaBoost" of Xml Inifile
                        TiXmlElement *p_xmlElement_Root_OD = NULL;
                        p_xmlElement_Root_OD = p_xmlElement_Root->FirstChildElement("adaBoost");

                        if (p_xmlElement_Root_OD)
                        {

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Faces_increase_search_scale
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                TiXmlElement *p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Faces_increase_search_scale");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_faces_increase_search_scale) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Faces_increase_search_scale'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Faces_increase_search_scale'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Faces_drop_groups
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Faces_drop_groups");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_faces_drop_groups) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Faces_drop_groups'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Faces_drop_groups'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Faces_min_search_scale_x
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Faces_min_search_scale_x");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_faces_min_search_scale_x) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Faces_min_search_scale_x'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Faces_min_search_scale_x'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Faces_min_search_scale_y
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Faces_min_search_scale_y");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_faces_min_search_scale_y) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Faces_min_search_scale_y'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Faces_min_search_scale_y'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Range_increase_search_scale
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Range_increase_search_scale");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_range_increase_search_scale) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Range_increase_search_scale'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Range_increase_search_scale'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Range_drop_groups
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Range_drop_groups");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_range_drop_groups) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Range_drop_groups'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Range_drop_groups'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Range_min_search_scale_x
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Range_min_search_scale_x");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_range_min_search_scale_x) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Range_min_search_scale_x'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Range_min_search_scale_x'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }

                                //************************************************************************************
                                //      BEGIN PeopleDetector->adaBoost->Range_min_search_scale_y
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Range_min_search_scale_y");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &people_detector_->m_range_min_search_scale_y) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Range_min_search_scale_y'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Range_min_search_scale_y'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }
                        }
                        //************************************************************************************
                        //      END CameraDataViewerControlFlow->adaBoost
                        //************************************************************************************

                        //************************************************************************************
                        //      BEGIN PeopleDetector->eigenfaces
                        //************************************************************************************
                        // Tag element "adaBoost" of Xml Inifile
                        p_xmlElement_Root_OD = NULL;
                        p_xmlElement_Root_OD = p_xmlElement_Root->FirstChildElement("eigenfaces");

                        if (p_xmlElement_Root_OD)
                        {

                                //************************************************************************************
                                //      BEGIN PeopleDetector->eigenfaces->Threshold_Face_Class
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                TiXmlElement *p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Threshold_Face_Class");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &threshold_) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Threshold_Face_Class'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Threshold_Face_Class'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }
                                //************************************************************************************
                                //      BEGIN PeopleDetector->eigenfaces->Threshold_Facespace
                                //************************************************************************************
                                // Subtag element "adaBoost" of Xml Inifile
                                p_xmlElement_Child = NULL;
                                p_xmlElement_Child = p_xmlElement_Root_OD->FirstChildElement("Threshold_Facespace");

                                if (p_xmlElement_Child)
                                {
                                        // read and save value of attribute
                                        if (p_xmlElement_Child->QueryValueAttribute("value", &threshold_fs_) != TIXML_SUCCESS)
                                        {
                                                std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                                std::cerr << "\t ... Can't find attribute 'value' of tag 'Threshold_Facespace'" << std::endl;
                                                return ipa_Utils::RET_FAILED;
                                        }
                                }
                                else
                                {
                                        std::cerr << "ERROR - PeopleDetector::LoadParameters:" << std::endl;
                                        std::cerr << "\t ... Can't find tag 'Threshold_Facespace'" << std::endl;
                                        return ipa_Utils::RET_FAILED;
                                }
                                //************************************************************************************
                                //      END CameraDataViewerControlFlow->eigenfaces
                                //************************************************************************************

                        }
                        else
                        {
                                std::cerr << "ERROR - CameraDataViewerControlFlow::LoadParameters:" << std::endl;
                                std::cerr << "\t ... Can't find tag 'ObjectDetectorParameters'" << std::endl;
                                return ipa_Utils::RET_FAILED;
                        }

                }

                //************************************************************************************
                //      END ObjectDetector
                //************************************************************************************
                else
                {
                        std::cerr << "ERROR - CameraDataViewerControlFlow::LoadParameters:" << std::endl;
                        std::cerr << "\t ... Can't find tag 'ObjectDetector'" << std::endl;
                        return ipa_Utils::RET_FAILED;
                }
        }
        return ipa_Utils::RET_OK;
}

//#######################
//#### main programm ####
//int main(int argc, char** argv)
//{
//      // Initialize ROS, specify name of node
//      ros::init(argc, argv, "people_detection");
//
//      // Create a handle for this node, initialize node
//      ros::NodeHandle nh;
//
//      // Create people detection node class instance
//      CobFaceDetectionNodelet people_detection_node(nh);
//
//      // Initialize people detection node
//      if (people_detection_node.init() != ipa_Utils::RET_OK)
//              return 0;
//
//      // Create action nodes
//      //DetectObjectsAction detect_action_node(object_detection_node, nh);
//      //AcquireObjectImageAction acquire_image_node(object_detection_node, nh);
//      //TrainObjectAction train_object_node(object_detection_node, nh);
//
//      ros::spin();
//
//      return 0;
//}