spin_image.h

/*
 * Software License Agreement (BSD License)
 *
 *  Point Cloud Library (PCL) - www.pointclouds.org
 *  Copyright (c) 2010-2011, Willow Garage, Inc.
 *  Copyright (c) 2012-, Open Perception, Inc.
 *
 *  All rights reserved.
 *
 *  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 copyright holder(s) nor the names of its
 *     contributors may be used to endorse or promote products derived
 *     from this software without specific prior written permission.
 *
 *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
 *  FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
 *  COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
 *  INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 *  BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 *  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
 *  CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 *  LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
 *  ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 *  POSSIBILITY OF SUCH DAMAGE.
 *
 * $Id$
 */
 
#pragma once
 
#include <pcl/point_types.h>
#include <pcl/features/feature.h>
 
namespace pcl
{
  /** \brief Estimates spin-image descriptors in the  given input points. 
    *  
    * This class represents spin image descriptor. Spin image is
    * a histogram of point locations summed along the bins of the image.
    * A 2D accumulator indexed by <VAR>a</VAR> and <VAR>b</VAR> is created. Next, 
    * the coordinates (<VAR>a</VAR>, <VAR>b</VAR>) are computed for a vertex in 
    * the surface mesh that is within the support of the spin image 
    * (explained below). The bin indexed by (<VAR>a</VAR>, <VAR>b</VAR>) in 
    * the accumulator is then incremented; bilinear interpolation is used 
    * to smooth the contribution of the vertex. This procedure is repeated 
    * for all vertices within the support of the spin image. 
    * The resulting accumulator can be thought of as an image; 
    * dark areas in the image correspond to bins that contain many projected points. 
    * As long as the size of the bins in the accumulator is greater 
    * than the median distance between vertices in the mesh 
    * (the definition of mesh resolution), the position of individual 
    * vertices will be averaged out during spin image generation.
    *
    * \attention The input normals given by \ref setInputNormals have to match
    * the input point cloud given by \ref setInputCloud. This behavior is
    * different than feature estimation methods that extend \ref
    * FeatureFromNormals, which match the normals with the search surface.
    *
    * With the default parameters, pcl::Histogram<153> is a good choice for PointOutT.
    * Of course the dimension of this descriptor must change to match the number
    * of bins set by the parameters.
    *
    * For further information please see:
    *
    *  - Johnson, A. E., & Hebert, M. (1998). Surface Matching for Object
    *    Recognition in Complex 3D Scenes. Image and Vision Computing, 16,
    *    635-651.
    * 
    * The class also implements radial spin images and spin-images in angular domain 
    * (or both).
    * 
    * \author Roman Shapovalov, Alexander Velizhev
    * \ingroup features
    */
  template <typename PointInT, typename PointNT, typename PointOutT>
  class SpinImageEstimation : public Feature<PointInT, PointOutT>
  {
    public:
      using Ptr = shared_ptr<SpinImageEstimation<PointInT, PointNT, PointOutT> >;
      using ConstPtr = shared_ptr<const SpinImageEstimation<PointInT, PointNT, PointOutT> >;
      using Feature<PointInT, PointOutT>::feature_name_;
      using Feature<PointInT, PointOutT>::getClassName;
      using Feature<PointInT, PointOutT>::indices_;
      using Feature<PointInT, PointOutT>::search_radius_;
      using Feature<PointInT, PointOutT>::k_;
      using Feature<PointInT, PointOutT>::surface_;
      using Feature<PointInT, PointOutT>::fake_surface_;
      using PCLBase<PointInT>::input_;
 
      using PointCloudOut = typename Feature<PointInT, PointOutT>::PointCloudOut;
 
      using PointCloudN = pcl::PointCloud<PointNT>;
      using PointCloudNPtr = typename PointCloudN::Ptr;
      using PointCloudNConstPtr = typename PointCloudN::ConstPtr;
 
      using PointCloudIn = pcl::PointCloud<PointInT>;
      using PointCloudInPtr = typename PointCloudIn::Ptr;
      using PointCloudInConstPtr = typename PointCloudIn::ConstPtr;
      
      /** \brief Constructs empty spin image estimator.
        * 
        * \param[in] image_width spin-image resolution, number of bins along one dimension
        * \param[in] support_angle_cos minimal allowed cosine of the angle between 
        *   the normals of input point and search surface point for the point 
        *   to be retained in the support
        * \param[in] min_pts_neighb min number of points in the support to correctly estimate 
        *   spin-image. If at some point the support contains less points, exception is thrown
        */
      SpinImageEstimation (unsigned int image_width = 8,
                           double support_angle_cos = 0.0,   // when 0, this is bogus, so not applied
                           unsigned int min_pts_neighb = 0);
      
      /** \brief Empty destructor */
      ~SpinImageEstimation () {}
 
      /** \brief Sets spin-image resolution.
        * 
        * \param[in] bin_count spin-image resolution, number of bins along one dimension
        */
      void 
      setImageWidth (unsigned int bin_count)
      {
        image_width_ = bin_count;
      }
 
      /** \brief Sets the maximum angle for the point normal to get to support region.
        * 
        * \param[in] support_angle_cos minimal allowed cosine of the angle between 
        *   the normals of input point and search surface point for the point 
        *   to be retained in the support
        */
      void 
      setSupportAngle (double support_angle_cos)
      {
        if (0.0 > support_angle_cos || support_angle_cos > 1.0)  // may be permit negative cosine?
        {
          throw PCLException ("Cosine of support angle should be between 0 and 1",
            "spin_image.h", "setSupportAngle");
        }
 
        support_angle_cos_ = support_angle_cos;
      }
 
      /** \brief Sets minimal points count for spin image computation.
        *
        * \param[in] min_pts_neighb min number of points in the support to correctly estimate 
        *   spin-image. If at some point the support contains less points, exception is thrown
        */
      void 
      setMinPointCountInNeighbourhood (unsigned int min_pts_neighb)
      {
        min_pts_neighb_ = min_pts_neighb;
      }
 
      /** \brief Provide a pointer to the input dataset that contains the point normals of 
        * the input XYZ dataset given by \ref setInputCloud
        * 
        * \attention The input normals given by \ref setInputNormals have to match
        * the input point cloud given by \ref setInputCloud. This behavior is
        * different than feature estimation methods that extend \ref
        * FeatureFromNormals, which match the normals with the search surface.
        * \param[in] normals the const boost shared pointer to a PointCloud of normals. 
        * By convention, L2 norm of each normal should be 1. 
        */
      inline void 
      setInputNormals (const PointCloudNConstPtr &normals)
      { 
        input_normals_ = normals; 
      }
 
      /** \brief Sets single vector a rotation axis for all input points.
        * 
        * It could be useful e.g. when the vertical axis is known.
        * \param[in] axis unit-length vector that serves as rotation axis for reference frame
        */
      void 
      setRotationAxis (const PointNT& axis)
      {
        rotation_axis_ = axis;
        use_custom_axis_ = true;
        use_custom_axes_cloud_ = false;
      }
 
      /** \brief Sets array of vectors as rotation axes for input points.
        * 
        * Useful e.g. when one wants to use tangents instead of normals as rotation axes
        * \param[in] axes unit-length vectors that serves as rotation axes for 
        *   the corresponding input points' reference frames
        */
      void 
      setInputRotationAxes (const PointCloudNConstPtr& axes)
      {
        rotation_axes_cloud_ = axes;
 
        use_custom_axes_cloud_ = true;
        use_custom_axis_ = false;
      }
 
      /** \brief Sets input normals as rotation axes (default setting). */
      void 
      useNormalsAsRotationAxis () 
      { 
        use_custom_axis_ = false; 
        use_custom_axes_cloud_ = false;
      }
 
      /** \brief Sets/unsets flag for angular spin-image domain.
        * 
        * Angular spin-image differs from the vanilla one in the way that not 
        * the points are collected in the bins but the angles between their
        * normals and the normal to the reference point. For further
        * information please see 
        * Endres, F., Plagemann, C., Stachniss, C., & Burgard, W. (2009). 
        * Unsupervised Discovery of Object Classes from Range Data using Latent Dirichlet Allocation. 
        * In Robotics: Science and Systems. Seattle, USA.
        * \param[in] is_angular true for angular domain, false for point domain
        */
      void 
      setAngularDomain (bool is_angular = true) { is_angular_ = is_angular; }
 
      /** \brief Sets/unsets flag for radial spin-image structure.
        * 
        * Instead of rectangular coordinate system for reference frame 
        * polar coordinates are used. Binning is done depending on the distance and 
        * inclination angle from the reference point
        * \param[in] is_radial true for radial spin-image structure, false for rectangular
        */
      void 
      setRadialStructure (bool is_radial = true) { is_radial_ = is_radial; }
 
    protected:
      /** \brief Estimate the Spin Image descriptors at a set of points given by
        * setInputWithNormals() using the surface in setSearchSurfaceWithNormals() and the spatial locator 
        * \param[out] output the resultant point cloud that contains the Spin Image feature estimates
        */
      void 
      computeFeature (PointCloudOut &output) override; 
 
      /** \brief initializes computations specific to spin-image.
        * 
        * \return true iff input data and initialization are correct
        */
      bool
      initCompute () override;
 
      /** \brief Computes a spin-image for the point of the scan. 
        * \param[in] index the index of the reference point in the input cloud
        * \return estimated spin-image (or its variant) as a matrix
        */
      Eigen::ArrayXXd 
      computeSiForPoint (int index) const;
 
    private:
      PointCloudNConstPtr input_normals_;
      PointCloudNConstPtr rotation_axes_cloud_;
      
      bool is_angular_;
 
      PointNT rotation_axis_;
      bool use_custom_axis_;
      bool use_custom_axes_cloud_;
 
      bool is_radial_;
 
      unsigned int image_width_;
      double support_angle_cos_;
      unsigned int min_pts_neighb_;
  };
}
 
#ifdef PCL_NO_PRECOMPILE
#include <pcl/features/impl/spin_image.hpp>
#endif