Estimates spin-image descriptors in the given input points. More...

#include <spin_image.h>

Inheritance diagram for pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >:

Public Types
typedef boost::shared_ptr < const SpinImageEstimation < PointInT, PointNT, PointOutT > >	ConstPtr
typedef pcl::PointCloud< PointInT >	PointCloudIn
typedef PointCloudIn::ConstPtr	PointCloudInConstPtr
typedef PointCloudIn::Ptr	PointCloudInPtr
typedef pcl::PointCloud< PointNT >	PointCloudN
typedef PointCloudN::ConstPtr	PointCloudNConstPtr
typedef PointCloudN::Ptr	PointCloudNPtr
typedef Feature< PointInT, PointOutT >::PointCloudOut	PointCloudOut
typedef boost::shared_ptr < SpinImageEstimation < PointInT, PointNT, PointOutT > >	Ptr
Public Member Functions
void	setAngularDomain (bool is_angular=true)
	Sets/unsets flag for angular spin-image domain.
void	setImageWidth (unsigned int bin_count)
	Sets spin-image resolution.
void	setInputNormals (const PointCloudNConstPtr &normals)
	Provide a pointer to the input dataset that contains the point normals of the input XYZ dataset given by setInputCloud.
void	setInputRotationAxes (const PointCloudNConstPtr &axes)
	Sets array of vectors as rotation axes for input points.
void	setMinPointCountInNeighbourhood (unsigned int min_pts_neighb)
	Sets minimal points count for spin image computation.
void	setRadialStructure (bool is_radial=true)
	Sets/unsets flag for radial spin-image structure.
void	setRotationAxis (const PointNT &axis)
	Sets single vector a rotation axis for all input points.
void	setSupportAngle (double support_angle_cos)
	Sets the maximum angle for the point normal to get to support region.
	SpinImageEstimation (unsigned int image_width=8, double support_angle_cos=0.0, unsigned int min_pts_neighb=0)
	Constructs empty spin image estimator.
void	useNormalsAsRotationAxis ()
	Sets input normals as rotation axes (default setting).
virtual	~SpinImageEstimation ()
	Empty destructor.
Protected Member Functions
virtual void	computeFeature (PointCloudOut &output)
	Estimate the Spin Image descriptors at a set of points given by setInputWithNormals() using the surface in setSearchSurfaceWithNormals() and the spatial locator.
Eigen::ArrayXXd	computeSiForPoint (int index) const
	Computes a spin-image for the point of the scan.
virtual bool	initCompute ()
	initializes computations specific to spin-image.
Private Attributes
unsigned int	image_width_
PointCloudNConstPtr	input_normals_
bool	is_angular_
bool	is_radial_
unsigned int	min_pts_neighb_
PointCloudNConstPtr	rotation_axes_cloud_
PointNT	rotation_axis_
double	support_angle_cos_
bool	use_custom_axes_cloud_
bool	use_custom_axis_

Detailed Description

template<typename PointInT, typename PointNT, typename PointOutT>
class pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >

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 a and b is created. Next, the coordinates (a, b) are computed for a vertex in the surface mesh that is within the support of the spin image (explained below). The bin indexed by (a, b) 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 setInputNormals have to match the input point cloud given by setInputCloud. This behavior is different than feature estimation methods that extend FeatureFromNormals, which match the normals with the search surface.

With the default paramters, 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

Definition at line 88 of file spin_image.h.

Member Typedef Documentation

template<typename PointInT, typename PointNT, typename PointOutT>

typedef boost::shared_ptr<const SpinImageEstimation<PointInT, PointNT, PointOutT> > pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::ConstPtr

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 92 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef pcl::PointCloud<PointInT> pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudIn

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 108 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef PointCloudIn::ConstPtr pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudInConstPtr

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 110 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef PointCloudIn::Ptr pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudInPtr

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 109 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef pcl::PointCloud<PointNT> pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudN

Definition at line 104 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef PointCloudN::ConstPtr pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudNConstPtr

Definition at line 106 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef PointCloudN::Ptr pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudNPtr

Definition at line 105 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef Feature<PointInT, PointOutT>::PointCloudOut pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::PointCloudOut

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 102 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

typedef boost::shared_ptr<SpinImageEstimation<PointInT, PointNT, PointOutT> > pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::Ptr

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 91 of file spin_image.h.

Constructor & Destructor Documentation

template<typename PointInT , typename PointNT , typename PointOutT >

pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::SpinImageEstimation	(	unsigned int	image_width = `8`,
		double	support_angle_cos = `0.0`,
		unsigned int	min_pts_neighb = `0`
	)

Constructs empty spin image estimator.

Parameters:

[in]	image_width	spin-image resolution, number of bins along one dimension
[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
[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

Definition at line 54 of file spin_image.hpp.

template<typename PointInT, typename PointNT, typename PointOutT>

virtual pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::~SpinImageEstimation ( ) [inline, virtual]

Empty destructor.

Definition at line 126 of file spin_image.h.

Member Function Documentation

template<typename PointInT , typename PointNT , typename PointOutT >

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::computeFeature ( PointCloudOut & output ) [protected, virtual]

Estimate the Spin Image descriptors at a set of points given by setInputWithNormals() using the surface in setSearchSurfaceWithNormals() and the spatial locator.

Parameters:

[out] output the resultant point cloud that contains the Spin Image feature estimates

Implements pcl::Feature< PointInT, PointOutT >.

Definition at line 324 of file spin_image.hpp.

template<typename PointInT , typename PointNT , typename PointOutT >

Eigen::ArrayXXd pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::computeSiForPoint ( int index ) const [protected]

Computes a spin-image for the point of the scan.

Parameters:

[in] index the index of the reference point in the input cloud

Returns:: estimated spin-image (or its variant) as a matrix

Definition at line 69 of file spin_image.hpp.

template<typename PointInT , typename PointNT , typename PointOutT >

bool pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::initCompute ( ) [protected, virtual]

initializes computations specific to spin-image.

Returns:: true iff input data and initialization are correct

Reimplemented from pcl::Feature< PointInT, PointOutT >.

Definition at line 239 of file spin_image.hpp.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setAngularDomain ( bool is_angular = true ) [inline]

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.

Parameters:

[in] is_angular true for angular domain, false for point domain

Definition at line 231 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setImageWidth ( unsigned int bin_count ) [inline]

Sets spin-image resolution.

Parameters:

[in] bin_count spin-image resolution, number of bins along one dimension

Definition at line 133 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setInputNormals ( const PointCloudNConstPtr & normals ) [inline]

Provide a pointer to the input dataset that contains the point normals of the input XYZ dataset given by setInputCloud.

Attention:: The input normals given by setInputNormals have to match the input point cloud given by setInputCloud. This behavior is different than feature estimation methods that extend FeatureFromNormals, which match the normals with the search surface.

Parameters:

[in] normals the const boost shared pointer to a PointCloud of normals. By convention, L2 norm of each normal should be 1.

Definition at line 178 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setInputRotationAxes ( const PointCloudNConstPtr & axes ) [inline]

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

Parameters:

[in] axes unit-length vectors that serves as rotation axes for the corresponding input points' reference frames

Definition at line 203 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setMinPointCountInNeighbourhood ( unsigned int min_pts_neighb ) [inline]

Sets minimal points count for spin image computation.

Parameters:

[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

Definition at line 162 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setRadialStructure ( bool is_radial = true ) [inline]

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

Parameters:

[in] is_radial true for radial spin-image structure, false for rectangular

Definition at line 241 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setRotationAxis ( const PointNT & axis ) [inline]

Sets single vector a rotation axis for all input points.

It could be useful e.g. when the vertical axis is known.

Parameters:

[in] axis unit-length vector that serves as rotation axis for reference frame

Definition at line 189 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::setSupportAngle ( double support_angle_cos ) [inline]

Sets the maximum angle for the point normal to get to support region.

Parameters:

[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

Definition at line 145 of file spin_image.h.

template<typename PointInT, typename PointNT, typename PointOutT>

void pcl::SpinImageEstimation< PointInT, PointNT, PointOutT >::useNormalsAsRotationAxis ( ) [inline]