update svn to r733 (3)

include/segmentation/FHGraph.h

@@ -0,0 +1,50 @@
+/**
+  * Point Cloud Segmentation using Felzenszwalb-Huttenlocher Algorithm
+  *
+  * Copyright (C) Jacobs University Bremen
+  *
+  * Released under the GPL version 3.
+  *
+  * @author Mihai-Cotizo Sima
+  */
+
+
+#ifndef __FHGRAPH_H_
+#define __FHGRAPH_H_
+
+#include <vector>
+#include <list>
+
+#include <slam6d/point.h>
+#include <slam6d/scan.h>
+#include <segmentation/segment-graph.h>
+#include <ANN/ANN.h>
+
+class FHGraph {
+public:
+    FHGraph(std::vector< Point >& ps, double weight(Point, Point), double sigma, double eps, int neighbors, float radius);
+    edge* getGraph();
+    Point operator[](int index);
+    int getNumPoints();
+    int getNumEdges();
+
+    void dispose();
+private:
+    void compute_neighbors(double weight(Point, Point), double eps);
+    void do_gauss(double sigma);
+    void without_gauss();
+
+    std::vector<edge> edges;
+    std::vector<Point>& points;
+    int V;
+    int E;
+
+    int nr_neighbors;
+    float radius;
+
+
+    struct he{ int x; float w; };
+    std::vector< std::list<he> > adjency_list;
+};
+
+#endif

include/segmentation/disjoint-set.h

@@ -0,0 +1,48 @@
+/*
+Copyright (C) 2006 Pedro Felzenszwalb
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 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 General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program; if not, write to the Free Software
+Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+*/
+
+#ifndef DISJOINT_SET
+#define DISJOINT_SET
+
+// disjoint-set forests using union-by-rank and path compression (sort of).
+
+typedef struct {
+    int rank;
+    int p;
+    int size;
+} uni_elt;
+
+class universe {
+public:
+    universe(int elements);
+    ~universe();
+    int find(int x);
+    void join(int x, int y);
+    int size(int x) const {
+        return elts[x].size;
+    }
+    int num_sets() const {
+        return num;
+    }
+
+private:
+    uni_elt *elts;
+    int num;
+};
+
+#endif

include/segmentation/segment-graph.h

@@ -0,0 +1,49 @@
+/*
+Copyright (C) 2006 Pedro Felzenszwalb
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 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 General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program; if not, write to the Free Software
+Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+*/
+
+#ifndef SEGMENT_GRAPH
+#define SEGMENT_GRAPH
+
+#include <algorithm>
+#include <cmath>
+#include "disjoint-set.h"
+
+// threshold function
+#define THRESHOLD(size, c) (c/size)
+
+typedef struct {
+    float w;
+    int a, b;
+} edge;
+
+bool operator<(const edge &a, const edge &b);
+
+/*
+ * Segment a graph
+ *
+ * Returns a disjoint-set forest representing the segmentation.
+ *
+ * num_vertices: number of vertices in graph.
+ * num_edges: number of edges in graph
+ * edges: array of edges.
+ * c: constant for treshold function.
+ */
+universe *segment_graph(int num_vertices, int num_edges, edge *edges,
+                        float c);
+
+#endif

include/slam6d/kdTreeImpl.h

@@ -0,0 +1,238 @@
+/** @file 
+ *  @brief Representation of the optimized k-d tree.
+ *  @author Remus Dumitru. Jacobs University Bremen, Germany
+ *  @author Corneliu-Claudiu Prodescu. Jacobs University Bremen, Germany
+ *  @author Andreas Nuechter. Institute of Computer Science, University of Osnabrueck, Germany.
+ *  @author Kai Lingemann. Institute of Computer Science, University of Osnabrueck, Germany.
+ *  @author Thomas Escher. Institute of Computer Science, University of Osnabrueck, Germany.
+ */
+
+#ifndef __KD_TREE_IMPL_H__
+#define __KD_TREE_IMPL_H__
+
+#include "slam6d/kdparams.h"
+#include "globals.icc"
+
+#ifdef _MSC_VER
+#if !defined _OPENMP && defined OPENMP 
+#define _OPENMP
+#endif
+#endif
+
+#ifdef _OPENMP
+#include <omp.h>
+#endif
+
+/**
+ * @brief The optimized k-d tree. 
+ * 
+ * A kD tree for points, with limited
+ * capabilities (find nearest point to
+ * a given point, or to a ray).
+ **/
+template<class PointData, class AccessorData, class AccessorFunc>
+class KDTreeImpl {
+public:
+  inline KDTreeImpl() { }
+  
+  virtual inline ~KDTreeImpl() {
+    if (!npts) {
+#ifdef WITH_OPENMP_KD
+      omp_set_num_threads(OPENMP_NUM_THREADS);
+#pragma omp parallel for schedule(dynamic)
+#endif
+      for (int i = 0; i < 2; i++) {
+        if (i == 0 && node.child1) delete node.child1;
+        if (i == 1 && node.child2) delete node.child2;
+      }
+    } else {
+      if (leaf.p) delete [] leaf.p;
+    }
+  }
+
+  virtual void create(PointData pts, AccessorData *indices, size_t n) {
+    AccessorFunc point;
+    
+    // Find bbox
+    double xmin = point(pts, indices[0])[0], xmax = point(pts, indices[0])[0];
+    double ymin = point(pts, indices[0])[1], ymax = point(pts, indices[0])[1];
+    double zmin = point(pts, indices[0])[2], zmax = point(pts, indices[0])[2];
+    for(unsigned int i = 1; i < n; i++) {
+      xmin = min(xmin, point(pts, indices[i])[0]);
+      xmax = max(xmax, point(pts, indices[i])[0]);
+      ymin = min(ymin, point(pts, indices[i])[1]);
+      ymax = max(ymax, point(pts, indices[i])[1]);
+      zmin = min(zmin, point(pts, indices[i])[2]);
+      zmax = max(zmax, point(pts, indices[i])[2]);
+    }
+
+    // Leaf nodes
+    if ((n > 0) && (n <= 10)) {
+      npts = n;
+      leaf.p = new AccessorData[n];
+      // fill leaf index array with indices
+      for(unsigned int i = 0; i < n; ++i) {
+        leaf.p[i] = indices[i];
+      }
+      return;
+    }
+  
+    // Else, interior nodes
+    npts = 0;
+  
+    node.center[0] = 0.5 * (xmin+xmax);
+    node.center[1] = 0.5 * (ymin+ymax);
+    node.center[2] = 0.5 * (zmin+zmax);
+    node.dx = 0.5 * (xmax-xmin);
+    node.dy = 0.5 * (ymax-ymin);
+    node.dz = 0.5 * (zmax-zmin);
+    node.r2 = sqr(node.dx) + sqr(node.dy) + sqr(node.dz);
+
+    // Find longest axis
+    if (node.dx > node.dy) {
+      if (node.dx > node.dz) {
+        node.splitaxis = 0;
+      } else {
+        node.splitaxis = 2;
+      }
+    } else {
+      if (node.dy > node.dz) {
+        node.splitaxis = 1;
+      } else {
+        node.splitaxis = 2;
+      }
+    }
+
+    // Partition
+    double splitval = node.center[node.splitaxis];
+
+    if ( fabs(max(max(node.dx,node.dy),node.dz)) < 0.01 ) {
+      npts = n;
+      leaf.p = new AccessorData[n];
+      // fill leaf index array with indices
+      for(unsigned int i = 0; i < n; ++i) {
+        leaf.p[i] = indices[i];
+      }
+      return;
+    }
+
+    AccessorData* left = indices, * right = indices + n - 1;
+    while(true) {
+      while(point(pts, *left)[node.splitaxis] < splitval)
+        left++;
+      while(point(pts, *right)[node.splitaxis] >= splitval)
+        right--;
+      if(right < left)
+        break;
+      std::swap(*left, *right);
+    }
+  
+    // Build subtrees
+    int i;
+#ifdef WITH_OPENMP_KD                   // does anybody know the reason why this is slower ?? --Andreas
+    omp_set_num_threads(OPENMP_NUM_THREADS);
+#pragma omp parallel for schedule(dynamic) 
+#endif
+    for (i = 0; i < 2; i++) {
+      if (i == 0) {
+        node.child1 = new KDTreeImpl();
+        node.child1->create(pts, indices, left - indices);
+      }
+      if (i == 1) {
+        node.child2 = new KDTreeImpl();
+        node.child2->create(pts, left, n - (left - indices));
+      }
+    }
+  }
+
+protected:
+  /**
+   * storing the parameters of the k-d tree, i.e., the current closest point,
+   * the distance to the current closest point and the point itself.
+   * These global variable are needed in this search.
+   *
+   * Padded in the parallel case.
+   */
+#ifdef _OPENMP
+#ifdef __INTEL_COMPILER
+  __declspec (align(16)) static KDParams params[MAX_OPENMP_NUM_THREADS];
+#else
+  static KDParams params[MAX_OPENMP_NUM_THREADS];
+#endif //__INTEL_COMPILER
+#else
+  static KDParams params[MAX_OPENMP_NUM_THREADS];
+#endif	
+
+  /**
+   * number of points. If this is 0: intermediate node. If nonzero: leaf.
+   */
+  int npts;
+  
+  /**
+   * Cue the standard rant about anon unions but not structs in C++
+   */
+  union {
+    /** 
+     * in case of internal node... 
+     */
+    struct {	 
+      double center[3]; ///< storing the center of the voxel (R^3)
+      double dx,  ///< defining the voxel itself
+	     dy,  ///< defining the voxel itself
+	     dz,  ///< defining the voxel itself
+	     r2;  ///< defining the voxel itself
+      int splitaxis;   ///< defining the kind of splitaxis
+      KDTreeImpl *child1;  ///< pointers to the childs
+      KDTreeImpl *child2;  ///< pointers to the childs
+    } node;
+    /** 
+     * in case of leaf node ... 
+     */
+    struct {
+      /** 
+       * store the value itself
+       * Here we store just a pointer to the data
+       */
+      AccessorData* p;
+    } leaf;
+  };
+
+  void _FindClosest(const PointData& pts, int threadNum) const {
+    AccessorFunc point;
+
+    // Leaf nodes
+    if (npts) {
+      for (int i = 0; i < npts; i++) {
+        double myd2 = Dist2(params[threadNum].p, point(pts, leaf.p[i]));
+        if (myd2 < params[threadNum].closest_d2) {
+          params[threadNum].closest_d2 = myd2;
+          params[threadNum].closest = point(pts, leaf.p[i]);
+        }
+      }
+      return;
+    }
+
+    // Quick check of whether to abort  
+    double approx_dist_bbox = max(max(fabs(params[threadNum].p[0]-node.center[0])-node.dx,
+                               fabs(params[threadNum].p[1]-node.center[1])-node.dy),
+						       fabs(params[threadNum].p[2]-node.center[2])-node.dz);
+    if (approx_dist_bbox >= 0 && sqr(approx_dist_bbox) >= params[threadNum].closest_d2)
+      return;
+
+    // Recursive case
+    double myd = node.center[node.splitaxis] - params[threadNum].p[node.splitaxis];
+    if (myd >= 0.0) {
+      node.child1->_FindClosest(pts, threadNum);
+      if (sqr(myd) < params[threadNum].closest_d2) {
+        node.child2->_FindClosest(pts, threadNum);
+      }
+    } else {
+      node.child2->_FindClosest(pts, threadNum);
+      if (sqr(myd) < params[threadNum].closest_d2) {
+        node.child1->_FindClosest(pts, threadNum);
+      }
+    }
+  }
+};
+
+#endif

src/normals/CMakeLists.txt~

@@ -0,0 +1,7 @@
+IF(WITH_NORMALS)
+  FIND_PACKAGE(OpenCV REQUIRED)
+
+  add_executable(calculateNormals calculate_normals.cc ../slam6d/fbr/fbr_global.cc ../slam6d/fbr/panorama.cc)
+
+  target_link_libraries(calculateNormals scan ANN newmat fbr_panorama fbr_cv_io ${Boost_LIBRARIES} ${OpenCV_LIBS})
+ENDIF(WITH_NORMALS)

src/normals/calculate_normals.cc~

@@ -0,0 +1,643 @@
+/*
+ * calculateNormals implementation
+ *
+ * Copyright (C) Johannes Schauer, Razvan Mihaly
+ *
+ * Released under the GPL version 3
+ *
+ */
+
+#include "ANN/ANN.h"
+
+#include "newmat/newmat.h"
+#include "newmat/newmatap.h"
+#include "newmat/newmatio.h"
+using namespace NEWMAT;
+
+#include "slam6d/point.h"
+#include "normals/pointNeighbor.h"
+#include "slam6d/scan.h"
+#include "slam6d/globals.icc"
+#include "slam6d/fbr/panorama.h"
+
+#include "normals/point.h"
+#include "normals/SRI.h"
+
+#include <string>
+using std::string;
+
+#include <iostream>
+using std::cout;
+using std::endl;
+using std::vector;
+
+#include <algorithm>
+
+#include <boost/program_options.hpp>
+#include <boost/filesystem/operations.hpp>
+#include <boost/filesystem/fstream.hpp>
+namespace po = boost::program_options;
+
+enum normal_method {KNN_PCA, AKNN_PCA, PANO_PCA, PANO_SRI};
+
+void normal_option_dependency(const po::variables_map & vm, normal_method ntype, const char *option)
+{
+    if (vm.count("normalMethod") && vm["normalMethod"].as<normal_method>() == ntype) {
+        if (!vm.count(option)) {
+            throw std::logic_error (string("this normal method needs ")+option+" to be set");
+        }
+    }
+}
+
+void normal_option_conflict(const po::variables_map & vm, normal_method ntype, const char *option)
+{
+    if (vm.count("normalMethod") && vm["normalMethod"].as<normal_method>() == ntype) {
+        if (vm.count(option)) {
+            throw std::logic_error (string("this normal method is incompatible with ")+option);
+        }
+    }
+}
+
+/*
+ * validates input type specification
+ */
+void validate(boost::any& v, const std::vector<std::string>& values,
+        IOType*, int) {
+    if (values.size() == 0)
+        throw std::runtime_error("Invalid model specification");
+    string arg = values.at(0);
+    try {
+        v = formatname_to_io_type(arg.c_str());
+    } catch (...) { // runtime_error
+        throw std::runtime_error("Format " + arg + " unknown.");
+    }
+}
+
+void validate(boost::any& v, const std::vector<std::string>& values,
+        normal_method*, int) {
+    if (values.size() == 0)
+        throw std::runtime_error("Invalid model specification");
+    string arg = values.at(0);
+    if(strcasecmp(arg.c_str(), "KNN_PCA") == 0) v = KNN_PCA;
+    else if(strcasecmp(arg.c_str(), "AKNN_PCA") == 0) v = AKNN_PCA;
+    else if(strcasecmp(arg.c_str(), "PANO_PCA") == 0) v = PANO_PCA;
+    else if(strcasecmp(arg.c_str(), "PANO_SRI") == 0) v = PANO_SRI;
+    else throw std::runtime_error(std::string("normal method ") + arg + std::string(" is unknown"));
+}
+
+/*
+ * parse commandline options, fill arguments
+ */
+void parse_options(int argc, char **argv, int &start, int &end,
+        bool &scanserver, string &dir, IOType &iotype,
+        int &maxDist, int &minDist, normal_method &normalMethod, int &knn,
+        int &kmin, int &kmax, double& alpha, int &width, int &height,
+        bool &flipnormals, double &factor)
+{
+    po::options_description generic("Generic options");
+    generic.add_options()
+        ("help,h", "output this help message");
+
+    po::options_description input("Input options");
+    input.add_options()
+        ("start,s", po::value<int>(&start)->default_value(0),
+         "start at scan <arg> (i.e., neglects the first <arg> scans) "
+         "[ATTENTION: counting naturally starts with 0]")
+        ("end,e", po::value<int>(&end)->default_value(-1),
+         "end after scan <arg>")
+        ("format,f", po::value<IOType>(&iotype)->default_value(UOS),
+         "using shared library <arg> for input. (chose F from {uos, uos_map, "
+         "uos_rgb, uos_frames, uos_map_frames, old, rts, rts_map, ifp, "
+         "riegl_txt, riegl_rgb, riegl_bin, zahn, ply})")
+        ("max,M", po::value<int>(&maxDist)->default_value(-1),
+         "neglegt all data points with a distance larger than <arg> 'units")
+        ("min,m", po::value<int>(&minDist)->default_value(-1),
+         "neglegt all data points with a distance smaller than <arg> 'units")
+        ("scanserver,S", po::bool_switch(&scanserver),
+         "Use the scanserver as an input method and handling of scan data")
+        ;
+
+    po::options_description normal("Normal options");
+    normal.add_options()
+        ("normalMethod,N", po::value<normal_method>(&normalMethod)->default_value(KNN_PCA),
+         "choose the method for computing normals:\n"
+         "KNN_PCA  -- use kNN and PCA\n"
+         "AKNN_PCA -- use adaptive kNN and PCA\n"
+         "PANO_PCA -- use panorama image neighbors and PCA\n"
+         "PANO_SRI -- use panorama image neighbors and spherical range image differentiation\n")
+        ("knn,K", po::value<int>(&knn),
+         "select the k in kNN search")
+        ("kmin,1", po::value<int>(&kmin),
+         "select k_min in adaptive kNN search")
+        ("kmax,2", po::value<int>(&kmax),
+         "select k_max in adaptive kNN search")
+        ("alpha,a", po::value<double>(&alpha),
+         "select the alpha parameter for detecting an ill-conditioned neighborhood")
+        ("width,w", po::value<int>(&width),
+         "width of panorama")
+        ("height,h", po::value<int>(&height),
+         "height of panorama")
+        ("flipnormals,F", po::bool_switch(&flipnormals),
+         "flip orientation of normals towards scan pose")
+        ("factor,c", po::value<double>(&factor),
+         "factor for SRI computation")
+        ;
+
+    po::options_description hidden("Hidden options");
+    hidden.add_options()
+        ("input-dir", po::value<string>(&dir), "input dir");
+
+    // all options
+    po::options_description all;
+    all.add(generic).add(input).add(normal).add(hidden);
+
+    // options visible with --help
+    po::options_description cmdline_options;
+    cmdline_options.add(generic).add(input).add(normal);
+
+    // positional argument
+    po::positional_options_description pd;
+    pd.add("input-dir", 1);
+
+    // process options
+    po::variables_map vm;
+    po::store(po::command_line_parser(argc, argv).
+            options(all).positional(pd).run(), vm);
+    po::notify(vm);
+
+    // display help
+    if (vm.count("help")) {
+        cout << cmdline_options;
+        exit(0);
+    }
+
+    normal_option_dependency(vm, KNN_PCA, "knn");
+    normal_option_conflict(vm, KNN_PCA, "kmin");
+    normal_option_conflict(vm, KNN_PCA, "kmax");
+    normal_option_conflict(vm, KNN_PCA, "alpha");
+    normal_option_conflict(vm, KNN_PCA, "width");
+    normal_option_conflict(vm, KNN_PCA, "height");
+    normal_option_conflict(vm, KNN_PCA, "factor");
+
+    normal_option_conflict(vm, AKNN_PCA, "knn");
+    normal_option_dependency(vm, AKNN_PCA, "kmin");
+    normal_option_dependency(vm, AKNN_PCA, "kmax");
+    normal_option_dependency(vm, AKNN_PCA, "alpha");
+    normal_option_conflict(vm, AKNN_PCA, "width");
+    normal_option_conflict(vm, AKNN_PCA, "height");
+    normal_option_conflict(vm, AKNN_PCA, "factor");
+
+    //normal_option_conflict(vm, PANO_PCA, "knn");
+    normal_option_dependency(vm, KNN_PCA, "knn");
+    normal_option_conflict(vm, PANO_PCA, "kmin");
+    normal_option_conflict(vm, PANO_PCA, "kmax");
+    normal_option_conflict(vm, PANO_PCA, "alpha");
+    normal_option_dependency(vm, PANO_PCA, "width");
+    normal_option_dependency(vm, PANO_PCA, "height");
+    normal_option_conflict(vm, PANO_PCA, "factor");
+
+    normal_option_conflict(vm, PANO_SRI, "knn");
+    normal_option_conflict(vm, PANO_SRI, "kmin");
+    normal_option_conflict(vm, PANO_SRI, "kmax");
+    normal_option_conflict(vm, PANO_SRI, "alpha");
+    normal_option_conflict(vm, PANO_SRI, "width");
+    normal_option_conflict(vm, PANO_SRI, "height");
+    normal_option_dependency(vm, PANO_SRI, "factor");
+
+    // add trailing slash to directory if not present yet
+    if (dir[dir.length()-1] != '/') dir = dir + "/";
+}
+
+/*
+ * retrieve a cv::Mat with x,y,z,r from a scan object
+ * functionality borrowed from scan_cv::convertScanToMat but this function
+ * does not allow a scanserver to be used, prints to stdout and can only
+ * handle a single scan
+ */
+void scan2mat(Scan* scan, cv::Mat& scan_cv) {
+    DataXYZ xyz = scan->get("xyz");
+    unsigned int nPoints = xyz.size();
+    scan_cv.create(nPoints,1,CV_32FC(4));
+    scan_cv = cv::Scalar::all(0);
+    cv::MatIterator_<cv::Vec3f> it = scan_cv.begin<cv::Vec3f>();
+    for(unsigned int i = 0; i < nPoints; i++){
+        (*it)[0] = xyz[i][0];
+        (*it)[1] = xyz[i][1];
+        (*it)[2] = xyz[i][2];
+        ++it;
+    }
+}
+
+/**
+ * Helper function that maps x, y, z to R, G, B using a linear function
+ */
+void mapNormalToRGB(const Point& normal, Point& rgb)
+{
+    rgb.x = 127.5 * normal.x + 127.5;
+    rgb.y = 127.5 * normal.y + 127.5;
+    rgb.z = 255.0 * fabs(normal.z);
+}
+
+/**
+ * Write normals to .3d files using the uos_rgb format
+ */
+void writeNormals(const Scan* scan, const string& dir,
+        const vector<Point>& points, const vector<Point>& normals)
+{
+
+    stringstream ss;
+    ss << dir << "scan" << string(scan->getIdentifier()) << ".3d";
+    ofstream scan_file;
+    scan_file.open(ss.str().c_str());
+    for(size_t i = 0;  i < points.size(); ++i) {
+        Point rgb;
+        mapNormalToRGB(normals[i], rgb);
+        scan_file
+            << points[i].x << " " << points[i].y << " " << points[i].z << " "
+            << (unsigned int) rgb.x << " " << (unsigned int) rgb.y << " "
+            << (unsigned int) rgb.z << "\n";
+    }
+    scan_file.close();
+
+    ss.clear(); ss.str(string());
+    ss << dir << "scan" << string(scan->getIdentifier()) << ".pose";
+    ofstream pose_file;
+    pose_file.open(ss.str().c_str());
+    pose_file << 0 << " " << 0 << " " << 0 << "\n" << 0 << " " << 0 << " " << 0 << "\n";
+    pose_file.close();
+}
+
+/**
+ * Compute eigen decomposition of a point and its neighbors using the NEWMAT library
+ * @param point - input points with corresponding neighbors
+ * @param e_values - out parameter returns the eigenvalues
+ * @param e_vectors - out parameter returns the eigenvectors
+ */
+void computeEigenDecomposition(const PointNeighbor& point,
+        DiagonalMatrix& e_values, Matrix& e_vectors)
+{
+    Point centroid(0, 0, 0);
+    vector<Point> neighbors = point.neighbors;
+
+    for (size_t j = 0; j < neighbors.size(); ++j) {
+        centroid.x += neighbors[j].x;
+        centroid.y += neighbors[j].y;
+        centroid.z += neighbors[j].z;
+    }
+    centroid.x /= (double) neighbors.size();
+    centroid.y /= (double) neighbors.size();
+    centroid.z /= (double) neighbors.size();
+
+    Matrix S(3, 3);
+    S = 0.0;
+    for (size_t j = 0; j < neighbors.size(); ++j) {
+        ColumnVector point_prime(3);
+        point_prime(1) = neighbors[j].x - centroid.x;
+        point_prime(2) = neighbors[j].y - centroid.y;
+        point_prime(3) = neighbors[j].z - centroid.z;
+        S = S + point_prime * point_prime.t();
+    }
+    // normalize S
+    for (int j = 0; j < 3; ++j)
+        for (int k = 0; k < 3; ++k)
+            S(j+1, k+1) /= (double) neighbors.size();
+
+    SymmetricMatrix C;
+    C << S;
+    // the decomposition
+    Jacobi(C, e_values, e_vectors);
+
+#ifdef DEBUG
+    // Print the result
+    cout << "The eigenvalues matrix:" << endl;
+    cout << e_values << endl;
+#endif
+}
+
+/**
+ * Compute neighbors using kNN search
+ * @param points - input set of points
+ * @param points_neighbors - output set of points with corresponding neighbors
+ * @param knn - k constant in kNN search
+ * @param kmax - to be used in adaptive knn search as the upper bound on adapting the k constant, defaults to -1 for regular kNN search
+ * @param alpha - to be used in adaptive knn search for detecting ill-conditioned neighborhoods
+ * @param eps - parameter required by the ANN library in kNN search
+ */
+void computeKNearestNeighbors(const vector<Point>& points,
+        vector<PointNeighbor>& points_neighbors, int knn, int kmax=-1,
+        double alpha=1000.0, double eps=1.0)
+{
+    ANNpointArray point_array = annAllocPts(points.size(), 3);
+    for (size_t i = 0; i < points.size(); ++i) {
+        point_array[i] = new ANNcoord[3];
+        point_array[i][0] = points[i].x;
+        point_array[i][1] = points[i].y;
+        point_array[i][2] = points[i].z;
+    }
+
+    ANNkd_tree t(point_array, points.size(), 3);
+    ANNidxArray n;
+    ANNdistArray d;
+
+    if (kmax < 0) {
+        /// regular kNN search, allocate memory for knn
+        n = new ANNidx[knn];
+        d = new ANNdist[knn];
+    } else {
+        /// adaptive kNN search, allocate memory for kmax
+        n = new ANNidx[kmax];
+        d = new ANNdist[kmax];
+    }
+
+    for (size_t i = 0; i < points.size(); ++i) {
+        vector<Point> neighbors;
+        ANNpoint p = point_array[i];
+
+        t.annkSearch(p, knn, n, d, eps);
+
+        neighbors.push_back(points[i]);
+        for (int j = 0; j < knn; ++j) {
+            if ( n[j] != (int)i )
+                neighbors.push_back(points[n[j]]);
+        }
+
+        PointNeighbor current_point(points[i], neighbors);
+        points_neighbors.push_back( current_point );
+        Matrix e_vectors(3,3); e_vectors = 0.0;
+        DiagonalMatrix e_values(3); e_values = 0.0;
+        computeEigenDecomposition( current_point, e_values, e_vectors );
+
+        if (kmax > 0) {
+            /// detecting an ill-conditioned neighborhood
+            if (e_values(3) / e_values(2) > alpha && e_values(2) > 0.0) {
+                if (knn < kmax)
+                    cout << "Increasing kmin to " << ++knn << endl;
+            }
+        }
+    }
+
+    delete[] n;
+    delete[] d;
+}
+
+/**
+ * Compute neighbors using kNN search
+ * @param point - input point with neighbors
+ * @param new_point - output point with new neighbors
+ * @param knn - k constant in kNN search
+ * @param eps - parameter required by the ANN library in kNN search
+ */
+void computeKNearestNeighbors(const PointNeighbor& point,
+        PointNeighbor& new_point, int knn, 
+        double eps=1.0) 
+{
+    /// allocate memory for all neighbors of point plus the point itself
+    ANNpointArray point_array = annAllocPts(point.neighbors.size()+1, 3);
+    for (size_t i = 0; i < point.neighbors.size(); ++i) {
+        point_array[i] = new ANNcoord[3];
+        point_array[i][0] = point.neighbors[i].x;
+        point_array[i][1] = point.neighbors[i].y;
+        point_array[i][2] = point.neighbors[i].z;
+    }
+    int last = point.neighbors.size();
+    point_array[last] = new ANNcoord[3];
+    point_array[last][0] = point.point.x;
+    point_array[last][1] = point.point.y;
+    point_array[last][2] = point.point.z;
+
+    ANNkd_tree t(point_array, point.neighbors.size()+1, 3);
+    ANNidxArray n;
+    ANNdistArray d;
+
+    /// regular kNN search, allocate memory for knn
+    n = new ANNidx[knn];
+    d = new ANNdist[knn];
+
+    vector<Point> new_neighbors;
+    /// last point in the array is the current point
+    ANNpoint p = point_array[point.neighbors.size()];
+
+    t.annkSearch(p, knn, n, d, eps);
+    new_neighbors.push_back(point.point);
+
+    for (int j = 0; j < knn; ++j) {
+        if ( n[j] != (int) point.neighbors.size() )
+            new_neighbors.push_back(point.neighbors[n[j]]);
+    }
+
+    new_point.point = point.point;
+    new_point.neighbors = new_neighbors;
+
+    delete[] n;
+    delete[] d;
+}
+
+/**
+ * Compute neighbors using panorama images
+ * @param fPanorama - input panorama image created from the current scan
+ * @param points_neighbors - output set of points with corresponding neighbors
+ */
+void computePanoramaNeighbors(Scan* scan,
+        vector<PointNeighbor>& points_neighbors, int width, int height, int knn)
+{
+    cv::Mat scan_cv;
+    scan2mat(scan, scan_cv);
+    fbr::panorama fPanorama(width, height, fbr::EQUIRECTANGULAR, 1, 0, fbr::EXTENDED);
+    fPanorama.createPanorama(scan_cv);
+    cv::Mat img = fPanorama.getRangeImage();
+    vector<vector<vector<cv::Vec3f> > > extended_map = fPanorama.getExtendedMap();
+    for (int row = 0; row < height; ++row) {
+        for (int col = 0; col < width; ++col) {
+            vector<cv::Vec3f> points_panorama = extended_map[row][col];
+            /// if no points found, skip pixel
+            if (points_panorama.size() < 1) continue;
+            /// foreach point from panorama consider all points in the bucket as its neighbors
+            for (size_t point_idx = 0; point_idx < points_panorama.size(); ++point_idx) {
+                Point point;
+                point.x = points_panorama[point_idx][0];
+                point.y = points_panorama[point_idx][1];
+                point.z = points_panorama[point_idx][2];
+                vector<Point> neighbors;
+                for (size_t i = 0; i < points_panorama.size(); ++i) {
+                    if (i != point_idx)
+                        neighbors.push_back(Point (points_panorama[i][0], points_panorama[i][1], points_panorama[i][2]) );
+                }
+                /// add neighbors from adjacent pixels and buckets
+                for (int i = -1; i <= 1; ++i) {
+                    for (int j = -1; j <= 1; ++j) {
+                        if (!(i==0 && j==0) && !(row+i < 0 || col+j < 0)
+                                &&  !(row+i >= height || col+j >= width) ) {
+                            vector<cv::Vec3f> neighbors_panorama = extended_map[row+i][col+j];
+                            for (size_t k = 0; k < neighbors_panorama.size(); ++k)
+                                neighbors.push_back(Point (neighbors_panorama[k][0],
+                                            neighbors_panorama[k][1],
+                                            neighbors_panorama[k][2]) );
+                        }
+                    }
+                }
+                /// filter the point by kNN search
+                PointNeighbor current_point(point, neighbors);
+                if (knn > 0) {
+                    PointNeighbor filtered_point;
+                    computeKNearestNeighbors(current_point, filtered_point, knn);
+                    points_neighbors.push_back(filtered_point);
+                } else {
+                    points_neighbors.push_back(current_point);
+                }
+            }
+        }
+    }
+}
+
+/**
+ * Compute normals using PCA given a set of points and their neighbors
+ * @param scan - pointer to current scan, used to compute the position vectors
+ * @param points - input set of points with corresponding neighbors
+ * @param normals - output set of normals
+ */
+void computePCA(const Scan* scan, const vector<PointNeighbor>& points,
+        vector<Point>& normals, bool flipnormals)
+{
+    ColumnVector origin(3);
+    const double *scan_pose = scan->get_rPos();
+    for (int i = 0; i < 3; ++i)
+        origin(i+1) = scan_pose[i];
+
+    for(size_t i = 0; i < points.size(); ++i) {
+        vector<Point> neighbors = points[i].neighbors;
+
+        if (points[i].neighbors.size() < 2) {
+            normals.push_back( Point(0,0,0) );
+            continue;
+        }
+
+        ColumnVector point_vector(3);
+        point_vector(1) = points[i].point.x - origin(1);
+        point_vector(2) = points[i].point.y - origin(2);
+        point_vector(3) = points[i].point.z - origin(3);
+        point_vector = point_vector / point_vector.NormFrobenius();
+
+        Matrix e_vectors(3,3); e_vectors = 0.0;
+        DiagonalMatrix e_values(3); e_values = 0.0;
+        computeEigenDecomposition(points[i], e_values, e_vectors);
+
+        ColumnVector v1(3);
+        v1(1) = e_vectors(1,1);
+        v1(2) = e_vectors(2,1);
+        v1(3) = e_vectors(3,1);
+        // consider first (smallest) eigenvector as the normal
+        Real angle = (v1.t() * point_vector).AsScalar();
+
+        // orient towards scan pose
+        // works better when orientation is not flipped
+        if (flipnormals && angle < 0) {
+            v1 *= -1.0;
+        }
+        normals.push_back( Point(v1(1), v1(2), v1(3)) );
+    }
+}
+
+void computeSRI(int factor, vector<Point>& points, vector<Point>& normals)
+{
+    SRI *sri2 = new SRI(0, factor);
+
+    for (size_t i = 0; i < points.size(); i++) {
+        sri2->addPoint(points[i].x, points[i].y, points[i].z);
+    }
+
+    points.clear();
+
+    for (unsigned int i = 0; i < sri2->points.size(); i++) {
+        double rgbN[3], x, y, z;
+        PointN* p = sri2->points[i];
+        p->getCartesian(x, y, z);
+        sri2->getNormalSRI(p, rgbN);
+        normals.push_back(Point(rgbN[0], rgbN[1], rgbN[2]));
+        points.push_back(Point(x, z, y));
+    }
+}
+
+void scan2points(Scan* scan, vector<Point> &points)
+{
+    DataXYZ xyz = scan->get("xyz");
+    unsigned int nPoints = xyz.size();
+    for(unsigned int i = 0; i < nPoints; ++i) {
+        points.push_back(Point(xyz[i][0], xyz[i][1], xyz[i][2]));
+    }
+}
+
+int main(int argc, char **argv)
+{
+    // commandline arguments
+    int start, end;
+    bool scanserver;
+    int maxDist, minDist;
+    string dir;
+    IOType iotype;
+    normal_method normalMethod;
+    int knn, kmin, kmax;
+    double alpha;
+    int width, height;
+    bool flipnormals;
+    double factor;
+
+    parse_options(argc, argv, start, end, scanserver, dir, iotype, maxDist,
+            minDist, normalMethod, knn, kmin, kmax, alpha, width, height,
+            flipnormals, factor);
+
+    Scan::openDirectory(scanserver, dir, iotype, start, end);
+
+    if(Scan::allScans.size() == 0) {
+        cerr << "No scans found. Did you use the correct format?" << endl;
+        exit(-1);
+    }
+
+    boost::filesystem::path boost_dir(dir + "normals/");
+    boost::filesystem::create_directory(boost_dir);
+
+    for(ScanVector::iterator it = Scan::allScans.begin(); it != Scan::allScans.end(); ++it) {
+        Scan* scan = *it;
+
+        // apply optional filtering
+        scan->setRangeFilter(maxDist, minDist);
+
+        vector<PointNeighbor> points_neighbors;
+        vector<Point> normals;
+        vector<Point> points;
+
+        scan2points(scan, points);
+
+        switch (normalMethod) {
+            case KNN_PCA:
+                computeKNearestNeighbors(points, points_neighbors, knn);
+                computePCA(scan, points_neighbors, normals, flipnormals);
+                break;
+            case AKNN_PCA:
+                computeKNearestNeighbors(points, points_neighbors, kmin, kmax, alpha);
+                computePCA(scan, points_neighbors, normals, flipnormals);
+                break;
+            case PANO_PCA:
+                computePanoramaNeighbors(scan, points_neighbors, width, height, knn);
+                computePCA(scan, points_neighbors, normals, flipnormals);
+                break;
+            case PANO_SRI:
+                computeSRI(factor, points, normals);
+                break;
+            default:
+                cerr << "unknown normal method" << endl;
+                return 1;
+                break;
+        }
+
+        if (points.size() != normals.size()) {
+            cerr << "got " << points.size() << " points but " << normals.size() << " normals" << endl;
+            return 1;
+        }
+
+        writeNormals(scan, dir + "normals/", points, normals);
+    }
+
+    Scan::closeDirectory();
+
+    return 0;
+}

src/normals/test.cc

@@ -0,0 +1,95 @@
+/*******************************************************
+ A simple program that demonstrates NewMat10 library.
+ The program defines a random symmetric matrix
+ and computes its eigendecomposition.
+ For further details read the NewMat10 Reference Manual
+********************************************************/
+
+#include <stdlib.h>
+#include <time.h>
+#include <string.h>
+
+// the following two are needed for printing
+#include <iostream.h>
+#include <iomanip.h>
+/**************************************
+/* The NewMat10 include files         */
+#include <include.h>
+#include <newmat.h>
+#include <newmatap.h>
+#include <newmatio.h>
+/***************************************/
+
+
+int main(int argc, char **argv) {
+	int M = 3, N = 5;
+	Matrix X(M,N); // Define an M x N general matrix
+
+	// Fill X by random numbers between 0 and 9
+	// Note that indexing into matrices in NewMat is 1-based!
+	srand(time(NULL));
+	for (int i = 1; i <= M; ++i) {
+		for (int j = 1; j <= N; ++j) {
+			X(i,j) = rand() % 10;
+		}
+	}
+
+	SymmetricMatrix C;
+	C << X * X.t(); // fill in C by X * X^t. 
+					// Works because we *know* that the result is symmetric
+
+	cout << "The symmetrix matrix C" << endl;
+	cout << setw(5) << setprecision(0) << C << endl;
+	
+
+	// compute eigendecomposition of C
+	Matrix			V(3,3); // for eigenvectors
+	DiagonalMatrix	D(3);   // for eigenvalues
+
+	// the decomposition
+	Jacobi(C, D, V);
+	
+	// Print the result
+	cout << "The eigenvalues matrix:" << endl;
+	cout << setw(10) << setprecision(5) << D << endl;
+	cout << "The eigenvectors matrix:" << endl;
+	cout << setw(10) << setprecision(5) << V << endl;
+
+	// Check that the first eigenvector indeed has the eigenvector property
+	ColumnVector v1(3);
+	v1(1) = V(1,1);
+	v1(2) = V(2,1);
+	v1(3) = V(3,1);
+
+	ColumnVector Cv1 = C * v1;
+	ColumnVector lambda1_v1 = D(1) * v1;
+
+	cout << "The max-norm of the difference between C*v1 and lambda1*v1 is " <<
+		   NormInfinity(Cv1 - lambda1_v1) << endl << endl;
+
+	// Build the inverse and check the result
+	Matrix Ci = C.i();
+	Matrix I  = Ci * C;
+
+	cout << "The inverse of C is" << endl;
+	cout << setw(10) << setprecision(5) << Ci << endl;
+	cout << "And the inverse times C is identity" << endl;
+	cout << setw(10) << setprecision(5) << I << endl;
+
+	// Example for multiple solves (see NewMat documentation)
+	ColumnVector r1(3), r2(3);
+	for (i = 1; i <= 3; ++i) {
+		r1(i) = rand() % 10;
+		r2(i) = rand() % 10;
+	}
+	LinearEquationSolver CLU = C; // decomposes C
+	ColumnVector s1 = CLU.i() * r1;
+	ColumnVector s2 = CLU.i() * r2;
+
+	cout << "solution for right hand side r1" << endl;
+	cout << setw(10) << setprecision(5) << s1 << endl;
+	cout << "solution for right hand side r2" << endl;
+	cout << setw(10) << setprecision(5) << s2 << endl;
+	
+	return 0;
+}

src/segmentation/FHGraph.cc

@@ -0,0 +1,269 @@
+/**
+  * Point Cloud Segmentation using Felzenszwalb-Huttenlocher Algorithm
+  *
+  * Copyright (C) Jacobs University Bremen
+  *
+  * Released under the GPL version 3.
+  *
+  * @author Mihai-Cotizo Sima
+  */
+
+#include <segmentation/FHGraph.h>
+#include <map>
+#include <omp.h>
+#include <algorithm>
+
+using namespace std;
+
+
+
+FHGraph::FHGraph(std::vector< Point >& ps, double weight(Point, Point), double sigma, double eps, int neighbors, float radius) :
+    points( ps ), V( ps.size() )
+{
+    /*
+     * 1. create adjency list using a map<int, vector<half_edge> >
+     * 2. use get_neighbors(e, max_dist) to get all the edges e' that are at a distance smaller than max_dist than e
+     * 3. using all these edges, compute the gaussian smoothed weight
+     * 4. insert the edges in a new list
+     */
+    nr_neighbors = neighbors;
+    this->radius = radius;
+
+    compute_neighbors(weight, eps);
+
+
+    if ( sigma > 0.01 )
+    {
+        do_gauss(sigma);
+    }
+    else
+    {
+        without_gauss();
+    }
+
+    adjency_list.clear();
+}
+
+void FHGraph::compute_neighbors(double weight(Point, Point), double eps)
+{
+
+    adjency_list.reserve(points.size());
+    adjency_list.resize(points.size());
+
+    ANNpointArray pa = annAllocPts(points.size(), 3);
+    for (size_t i=0; i<points.size(); ++i)
+    {
+        pa[i] = new ANNcoord[3];
+        pa[i][0] = points[i].x;
+        pa[i][1] = points[i].y;
+        pa[i][2] = points[i].z;
+    }
+
+    ANNkd_tree t(pa, points.size(), 3);
+
+    if ( radius < 0 ) // Using knn search
+    {
+        nr_neighbors++;
+        ANNidxArray n = new ANNidx[nr_neighbors];
+        ANNdistArray d = new ANNdist[nr_neighbors];
+
+        for (size_t i=0; i<points.size(); ++i)
+        {
+            ANNpoint p = pa[i];
+
+            t.annkSearch(p, nr_neighbors, n, d, eps);
+
+            for (int j=0; j<nr_neighbors; ++j)
+            {
+                if ( n[j] == (int)i ) continue;
+
+                he e;
+                e.x = n[j];
+                e.w = weight(points[i], points[n[j]]);
+
+                adjency_list[i].push_back(e);
+            }
+
+        }
+
+        delete[] n;
+        delete[] d;
+    }
+    else // Using radius search
+    {
+        float sqradius = radius*radius;
+
+
+        ANNidxArray n;
+        ANNdistArray d;
+        int nret;
+        int total = 0;
+
+        const int MOD = 1000;
+        int TMP = MOD;
+
+        for (size_t i=0; i<points.size(); ++i)
+        {
+            ANNpoint p = pa[i];
+
+            nret = t.annkFRSearch(p, sqradius, 0, NULL, NULL, eps);
+            total += nret;
+
+            n = new ANNidx[nret];
+            d = new ANNdist[nret];
+            t.annkFRSearch(p, sqradius, nret, n, d, eps);
+
+            if ( nr_neighbors > 0 && nr_neighbors < nret )
+            {
+                random_shuffle(n, n+nret);
+                nret = nr_neighbors;
+            }
+
+            for (int j=0; j<nret; ++j)
+            {
+                if ( n[j] == (int)i ) continue;
+
+                he e;
+                e.x = n[j];
+                e.w = weight(points[i], points[n[j]]);
+
+                adjency_list[i].push_back(e);
+            }
+
+            delete[] n;
+            delete[] d;
+            if ( TMP==0 )
+            {
+                TMP = MOD;
+                cout << "Point " << i << "/" << V << ", or "<< (i*100.0 / V) << "%\r"; cout.flush();
+            }
+            TMP --;
+        }
+        cout << "Average nr of neighbors: " << (float) total / points.size() << endl;
+
+    }
+
+    annDeallocPts(pa);
+}
+
+static double gauss(double x, double miu, double sigma)
+{
+    double tmp = ((x-miu)/sigma);
+    return exp(- .5 * tmp * tmp);
+}
+
+static void normalize(std::vector<double>& v)
+{
+    double s = 0;
+    for (size_t i=0; i<v.size(); ++i)
+        s += v[i];
+    for (size_t i=0; i<v.size(); ++i)
+        v[i] /= s;
+}
+
+string tostr(int x)
+{
+    stringstream ss;
+    ss << x;
+    return ss.str();
+}
+
+void FHGraph::do_gauss(double sigma)
+{
+    edges.reserve( V * nr_neighbors);
+    list<he>::iterator k, j;
+    vector<double> gauss_weight, edge_weight;
+    edge e;
+#pragma omp parallel for private(j, k, e, gauss_weight, edge_weight) schedule(dynamic)
+    for (int i=0; i<V; ++i)
+    {
+        for (j=adjency_list[i].begin();
+             j!=adjency_list[i].end();
+             j++)
+        {
+
+            gauss_weight.clear();
+            edge_weight.clear();
+
+            for (k=adjency_list[i].begin();
+                 k!=adjency_list[i].end();
+                 ++k)
+            {
+                gauss_weight.push_back(gauss(k->w, j->w, sigma));
+                edge_weight.push_back(k->w);
+            }
+            for (k=adjency_list[j->x].begin();
+                 k!=adjency_list[j->x].end();
+                 ++k)
+            {
+                gauss_weight.push_back(gauss(k->w, j->w, sigma));
+                edge_weight.push_back(k->w);
+            }
+            normalize(gauss_weight);
+
+            e.a = i; e.b = j->x;
+            e.w = 0;
+            for (size_t k=0; k<edge_weight.size(); ++k)
+                e.w += gauss_weight[k] * edge_weight[k];
+
+#pragma omp critical
+            {
+                edges.push_back(e);
+            }
+        }
+    }
+}
+
+void FHGraph::without_gauss()
+{
+    edges.reserve( V * nr_neighbors);
+    list<he>::iterator j;
+    edge e;
+
+    for (int i=0; i<V; ++i)
+    {
+        for (j=adjency_list[i].begin();
+             j!=adjency_list[i].end();
+             j++)
+        {
+            e.a = i; e.b = j->x; e.w = j->w;
+            edges.push_back(e);
+        }
+    }
+}
+
+edge* FHGraph::getGraph()
+{
+    edge* ret = new edge[edges.size()];
+    for (size_t i=0; i<edges.size(); ++i)
+        ret[i] = edges[i];
+    return ret;
+}
+
+Point FHGraph::operator[](int index)
+{
+    return points[index];
+}
+
+int FHGraph::getNumPoints()
+{
+    return V;
+}
+
+int FHGraph::getNumEdges()
+{
+    return edges.size();
+}
+
+template<typename T>
+void vectorFree(T& t) {
+    T tmp;
+    t.swap(tmp);
+}
+
+void FHGraph::dispose() {
+    vectorFree(edges);
+    vectorFree(points);
+    vectorFree(adjency_list);
+}
+

src/segmentation/disjoint-set.cc

@@ -0,0 +1,47 @@
+/**
+  * Point Cloud Segmentation using Felzenszwalb-Huttenlocher Algorithm
+  *
+  * Copyright (C) Jacobs University Bremen
+  *
+  * Released under the GPL version 3.
+  *
+  * @author Mihai-Cotizo Sima
+  */
+
+
+#include <segmentation/disjoint-set.h>
+
+universe::universe(int elements) {
+    elts = new uni_elt[elements];
+    num = elements;
+    for (int i = 0; i < elements; i++) {
+        elts[i].rank = 0;
+        elts[i].size = 1;
+        elts[i].p = i;
+    }
+}
+
+universe::~universe() {
+    delete [] elts;
+}
+
+int universe::find(int x) {
+    int y = x;
+    while (y != elts[y].p)
+        y = elts[y].p;
+    elts[x].p = y;
+    return y;
+}
+
+void universe::join(int x, int y) {
+    if (elts[x].rank > elts[y].rank) {
+        elts[y].p = x;
+        elts[x].size += elts[y].size;
+    } else {
+        elts[x].p = y;
+        elts[y].size += elts[x].size;
+        if (elts[x].rank == elts[y].rank)
+            elts[y].rank++;
+    }
+    num--;
+}

src/segmentation/fhsegmentation.cc

@@ -0,0 +1,297 @@
+/**
+  * Point Cloud Segmentation using Felzenszwalb-Huttenlocher Algorithm
+  *
+  * Copyright (C) Jacobs University Bremen
+  *
+  * Released under the GPL version 3.
+  *
+  * @author Billy Okal <b.okal@jacobs-university.de>
+  * @author Mihai-Cotizo Sima
+  * @file fhsegmentation.cc
+  */
+
+#include <iostream>
+#include <string>
+#include <fstream>
+#include <errno.h>
+
+#include <boost/program_options.hpp>
+
+#include <slam6d/io_types.h>
+#include <slam6d/globals.icc>
+#include <slam6d/scan.h>
+#include <scanserver/clientInterface.h>
+
+#include <segmentation/FHGraph.h>
+
+#ifdef _MSC_VER
+#define strcasecmp _stricmp
+#define strncasecmp _strnicmp
+#else
+#include <strings.h>
+#endif
+
+namespace po = boost::program_options;
+using namespace std;
+
+/// validate IO types
+void validate(boost::any& v, const std::vector<std::string>& values,
+    IOType*, int) {
+    if (values.size() == 0)
+        throw std::runtime_error("Invalid model specification");
+    string arg = values.at(0);
+    try {
+        v = formatname_to_io_type(arg.c_str());
+    } catch (...) { // runtime_error
+        throw std::runtime_error("Format " + arg + " unknown.");
+    }
+}
+
+/// Parse commandline options
+void parse_options(int argc, char **argv, int &start, int &end, bool &scanserver, int &max_dist, int &min_dist, string &dir,
+                   IOType &iotype, float &sigma, int &k, int &neighbors, float &eps, float &radius, int &min_size)
+{
+    /// ----------------------------------
+    /// set up program commandline options
+    /// ----------------------------------
+    po::options_description cmd_options("Usage: fhsegmentation <options> where options are (default values in brackets)");
+    cmd_options.add_options()
+            ("help,?", "Display this help message")
+            ("start,s", po::value<int>(&start)->default_value(0), "Start at scan number <arg>")
+            ("end,e", po::value<int>(&end)->default_value(-1), "Stop at scan number <arg>")
+            ("scanserver,S", po::value<bool>(&scanserver)->default_value(false), "Use the scanserver as an input method")
+            ("format,f", po::value<IOType>(&iotype)->default_value(UOS),
+             "using shared library <arg> for input. (chose format from [uos|uosr|uos_map|"
+             "uos_rgb|uos_frames|uos_map_frames|old|rts|rts_map|ifp|"
+             "riegl_txt|riegl_rgb|riegl_bin|zahn|ply])")
+            ("max,M", po::value<int>(&max_dist)->default_value(-1),"neglegt all data points with a distance larger than <arg> 'units")
+            ("min,m", po::value<int>(&min_dist)->default_value(-1), "neglegt all data points with a distance smaller than <arg> 'units")
+            ("K,k", po::value<int>(&k)->default_value(1), "<arg> value of K value used in the FH segmentation")
+            ("neighbors,N", po::value<int>(&neighbors)->default_value(1), "use approximate <arg>-nearest neighbors search or limit the number of points")
+            ("sigma,v", po::value<float>(&sigma)->default_value(1.0), "Set the Gaussian variance  for smoothing to <arg>")
+            ("radius,r", po::value<float>(&radius)->default_value(-1.0), "Set the range of radius search to <arg>")
+            ("eps,E", po::value<float>(&eps)->default_value(1.0), "Set error threshold used by the AKNN algorithm to <arg>")
+            ("minsize,z", po::value<int>(&min_size)->default_value(0), "Keep segments of size at least <arg>")
+            ;
+
+    po::options_description hidden("Hidden options");
+    hidden.add_options()
+        ("input-dir", po::value<string>(&dir), "input dir");
+
+    po::positional_options_description pd;
+    pd.add("input-dir", 1);
+
+    po::options_description all;
+    all.add(cmd_options).add(hidden);
+
+    po::variables_map vmap;
+    po::store(po::command_line_parser(argc, argv).options(all).positional(pd).run(), vmap);
+    po::notify(vmap);
+
+    if (vmap.count("help")) {
+        cout << cmd_options << endl;
+        exit(-1);
+    }
+
+    // read scan path
+    if (dir[dir.length()-1] != '/') dir = dir + "/";
+
+}
+
+/// distance measures
+double weight1(Point a, Point b)
+{
+    return a.distance(b);
+}
+
+double weight2(Point a, Point b)
+{
+    return a.distance(b) * .5 + fabs(a.reflectance-b.reflectance) * .5;
+}
+
+
+/// Write a pose file with the specofied name
+void writePoseFiles(string dir, const double* rPos, const double* rPosTheta, int num, int outnum)
+{
+    for (int i = outnum; i < num; i++) {
+        string poseFileName = dir + "segments/scan" + to_string(i, 3) + ".pose";
+        ofstream posout(poseFileName.c_str());
+
+        posout << rPos[0] << " "
+               << rPos[1] << " "
+               << rPos[2] << endl
+               << deg(rPosTheta[0]) << " "
+               << deg(rPosTheta[1]) << " "
+               << deg(rPosTheta[2]) << endl;
+        posout.clear();
+        posout.close();
+    }
+}
+
+/// write scan files for all segments
+void writeScanFiles(string dir, int outnum, const vector<vector<Point>* > cloud)
+{
+    for (int i = outnum, j = 0; i < (int)cloud.size() && j < (int)cloud.size(); i++, j++) {
+        vector<Point>* segment = cloud[j];
+        string scanFileName = dir + "segments/scan" + to_string(i,3) + ".3d";
+        ofstream scanout(scanFileName.c_str());
+
+        for (int k = 0; k < (int)segment->size(); k++) {
+            Point p = segment->at(k);
+            scanout << p.x << " " << p.y << " " << p.z << endl;
+        }
+        scanout.close();
+    }
+}
+
+
+
+/// =============================================
+/// Main
+/// =============================================
+int main(int argc, char** argv)
+{
+    int start, end;
+    bool scanserver;
+    int max_dist, min_dist;
+    string dir;
+    IOType iotype;
+    float sigma;
+    int k, neighbors;
+    float eps;
+    float radius;
+    int min_size;
+
+    parse_options(argc, argv, start, end, scanserver, max_dist, min_dist,
+                  dir, iotype, sigma, k, neighbors, eps, radius, min_size);
+
+    /// ----------------------------------
+    /// Prepare and read scans
+    /// ----------------------------------
+    if (scanserver) {
+        try {
+            ClientInterface::create();
+        } catch(std::runtime_error& e) {
+            cerr << "ClientInterface could not be created: " << e.what() << endl;
+            cerr << "Start the scanserver first." << endl;
+            exit(-1);
+        }
+    }
+
+    /// Make directory for saving the scan segments
+    string segdir = dir + "segments";
+
+#ifdef _MSC_VER
+    int success = mkdir(segdir.c_str());
+#else
+    int success = mkdir(segdir.c_str(), S_IRWXU|S_IRWXG|S_IRWXO);
+#endif
+    if(success == 0) {
+        cout << "Writing segments to " << segdir << endl;
+    } else if(errno == EEXIST) {
+        cout << "WARN: Directory " << segdir << " exists already. Contents will be overwriten" << endl;
+    } else {
+        cerr << "Creating directory " << segdir << " failed" << endl;
+        exit(1);
+    }
+
+    /// Read the scans
+    Scan::openDirectory(scanserver, dir, iotype, start, end);
+    if(Scan::allScans.size() == 0) {
+        cerr << "No scans found. Did you use the correct format?" << endl;
+        exit(-1);
+    }
+
+    /// --------------------------------------------
+    /// Initialize and perform segmentation
+    /// --------------------------------------------
+    std::vector<Scan*>::iterator it = Scan::allScans.begin();
+    int outscan = start;
+
+    for( ; it != Scan::allScans.end(); ++it) {
+        Scan* scan = *it;
+        const double* rPos = scan->get_rPos();
+        const double* rPosTheta = scan->get_rPosTheta();
+
+        /// read scan into points
+        DataXYZ xyz(scan->get("xyz"));
+        vector<Point> points;
+        points.reserve(xyz.size());
+
+        for(unsigned int j = 0; j < xyz.size(); j++) {
+            Point p(xyz[j][0], xyz[j][1], xyz[j][2]);
+            points.push_back(p);
+        }
+
+        /// create the graph and get the segments
+        cout << "creating graph" << endl;
+        FHGraph sgraph(points, weight2, sigma, eps, neighbors, radius);
+
+        cout << "segmenting graph" << endl;
+        edge* sedges = sgraph.getGraph();
+        universe* segmented = segment_graph(sgraph.getNumPoints(), 
+                                            sgraph.getNumEdges(), 
+                                            sedges, k);
+
+        cout << "post processing" << endl;
+        for (int i = 0; i < sgraph.getNumEdges(); ++i)
+        {
+            int a = sedges[i].a;
+            int b = sedges[i].b;
+
+            int aa = segmented->find(a);
+            int bb = segmented->find(b);
+
+            if ( (aa!=bb) && 
+                 (segmented->size(aa) < min_size || 
+                  segmented->size(bb) < min_size) )
+                segmented->join(aa, bb);
+        }
+
+        delete[] sedges;
+
+        int nr = segmented->num_sets();
+        cout << "Obtained " << nr << " segment(s)" << endl;
+
+        /// write point clouds with segments
+        vector< vector<Point>* > clouds;
+        clouds.reserve(nr);
+        for (int i=0; i<nr; ++i)
+            clouds.push_back( new vector<Point> );
+
+        map<int, int> components2cloud;
+        int kk = 0;
+
+        for (int i = 0; i < sgraph.getNumPoints(); ++i)
+        {
+            int component = segmented->find(i);
+            if ( components2cloud.find(component)==components2cloud.end() )
+            {
+                components2cloud[component] = kk++;
+                clouds[components2cloud[component]]->reserve(segmented->size(component));
+            }
+            clouds[components2cloud[component]]->push_back(sgraph[i]);
+        }
+
+        // pose file (repeated for the number of segments
+        writePoseFiles(dir, rPos, rPosTheta, clouds.size(), outscan);
+        // scan files for all segments
+        writeScanFiles(dir, outscan, clouds);
+
+        outscan += clouds.size();
+
+        /// clean up
+        sgraph.dispose();
+    }
+
+    // shutdown everything
+    if (scanserver)
+        ClientInterface::destroy();
+    else
+        Scan::closeDirectory();
+
+    cout << "Normal program end" << endl;
+
+    return 0;
+}
+

src/segmentation/segment-graph.cc

@@ -0,0 +1,50 @@
+/**
+  * Point Cloud Segmentation using Felzenszwalb-Huttenlocher Algorithm
+  *
+  * Copyright (C) Jacobs University Bremen
+  *
+  * Released under the GPL version 3.
+  *
+  * @author Mihai-Cotizo Sima
+  */
+
+#include <segmentation/segment-graph.h>
+
+bool operator<(const edge &a, const edge &b) {
+    return a.w < b.w;
+}
+
+universe *segment_graph(int num_vertices, int num_edges, edge *edges,
+                        float c) {
+    // sort edges by weight
+    std::sort(edges, edges + num_edges);
+
+    // make a disjoint-set forest
+    universe *u = new universe(num_vertices);
+
+    // init thresholds
+    float *threshold = new float[num_vertices];
+    for (int i = 0; i < num_vertices; i++)
+        threshold[i] = THRESHOLD(1,c);
+
+    // for each edge, in non-decreasing weight order...
+    for (int i = 0; i < num_edges; i++) {
+        edge *pedge = &edges[i];
+
+        // components conected by this edge
+        int a = u->find(pedge->a);
+        int b = u->find(pedge->b);
+        if (a != b) {
+            if ((pedge->w <= threshold[a]) &&
+                    (pedge->w <= threshold[b])) {
+                u->join(a, b);
+                a = u->find(a);
+                threshold[a] = pedge->w + THRESHOLD(u->size(a), c);
+            }
+        }
+    }
+
+    // free up
+    delete threshold;
+    return u;
+}