normals_estimator.hpp

/* $ Id: $*/

#ifndef NORMALS_ESTIMATOR_HPP
#define NORMALS_ESTIMATOR_HPP
#include "kernel/csvFile.hpp"
#include "jmath/jblas.hpp"
#include "jmath/pca.hpp"
#include "jmath/jann.hpp"
#include "kernel/jafarMacro.hpp"

namespace jafar {
  namespace modeler {

    class normals_estimator {
    private:
      size_t m_knn;
    public:
      ublas::matrix<double> m_cloud;
    private:
      jann::KD_tree_index< flann::L2<double> > m_index;
      jmath::PCA pca;
      size_t K;
      jblas::vec compute_normal(const ublas::matrix<double>& Q);

    public:
      normals_estimator(size_t k=25) : m_knn(k) {};

      normals_estimator(const ublas::matrix<double>& set, size_t k=25) : 
        m_knn(k), m_cloud (set), m_index(m_cloud, 4) {
        JFR_ASSERT(m_cloud.size2() == 3, "for now can handle only 3 dimension");
        JFR_ASSERT(m_cloud.size1() > m_knn, "nbr of points must exceed nbr of neighbours");
      }

      normals_estimator(const std::map<size_t, jblas::vec3>& points, 
                        size_t k = 25) : 
        m_knn(k), m_cloud(points.size(), 3) {
        JFR_ASSERT(points.size() > m_knn, "points size must be at least "<<m_knn);
        size_t i = 0;
        for(std::map<size_t, jblas::vec3>::const_iterator iter = points.begin(); 
            iter != points.end();
            ++iter, ++i) {
          row(m_cloud, i) = iter->second;
        }
        m_index(m_cloud, 4); 
      }

      jblas::vec estimate(int index);

      ublas::matrix<double> batch_estimate();
    };
  }
}

#endif