Point Cloud Library (PCL)  1.3.1
Public Types | Public Member Functions
pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget > Class Template Reference

GeneralizedIterativeClosestPoint is an ICP variant that implements the generalized iterative closest point algorithm as described by Alex Segal et al. More...

#include <pcl/registration/gicp.h>

Inheritance diagram for pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >:
Inheritance graph
[legend]
Collaboration diagram for pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >:
Collaboration graph
[legend]

List of all members.

Public Types

typedef boost::shared_ptr
< Registration< PointSource,
PointTarget > > 
Ptr
typedef boost::shared_ptr
< const Registration
< PointSource, PointTarget > > 
ConstPtr
typedef pcl::KdTree< PointTarget > KdTree
typedef pcl::KdTree
< PointTarget >::Ptr 
KdTreePtr
typedef
KdTree::PointRepresentationConstPtr 
PointRepresentationConstPtr
typedef
pcl::registration::TransformationEstimation
< PointSource, PointTarget > 
TransformationEstimation
typedef
TransformationEstimation::Ptr 
TransformationEstimationPtr
typedef
TransformationEstimation::ConstPtr 
TransformationEstimationConstPtr

Public Member Functions

 GeneralizedIterativeClosestPoint ()
 Empty constructor.
void setInputCloud (const PointCloudSourceConstPtr &cloud)
 Provide a pointer to the input dataset.
void setInputTarget (const PointCloudTargetConstPtr &target)
 Provide a pointer to the input target (e.g., the point cloud that we want to align the input source to)
void estimateRigidTransformationLM (const PointCloudSource &cloud_src, const PointCloudTarget &cloud_tgt, Eigen::Matrix4f &transformation_matrix)
 Estimate a rigid rotation transformation between a source and a target point cloud using an iterative non-linear Levenberg-Marquardt approach.
void estimateRigidTransformationLM (const PointCloudSource &cloud_src, const std::vector< int > &indices_src, const PointCloudTarget &cloud_tgt, const std::vector< int > &indices_tgt, Eigen::Matrix4f &transformation_matrix)
 Estimate a rigid rotation transformation between a source and a target point cloud using an iterative non-linear Levenberg-Marquardt approach.
const Eigen::Matrix3d & mahalanobis (size_t index) const
void computeRDerivative (const double x[], const Eigen::Matrix3d &R, double g[])
 Computes rotation matrix derivative.
void setRotationEpsilon (double epsilon)
 Set the rotation epsilon (maximum allowable difference between two consecutive rotations) in order for an optimization to be considered as having converged to the final solution.
double getRotationEpsilon ()
 Get the rotation epsilon (maximum allowable difference between two consecutive rotations) as set by the user.
void setCorrespondenceRandomness (int k)
 Set the number of neighbors used when selecting a point neighbourhood to compute covariances.
void getCorrespondenceRandomness ()
 Get the number of neighbors used when computing covariances as set by the user.
 GeneralizedIterativeClosestPoint ()
 Empty constructor.
void setMaxDistance (double max_distance)
void setTransformationEstimation (const TransformationEstimationPtr &te)
PointCloudTargetConstPtr const getInputTarget ()
 Get a pointer to the input point cloud dataset target.
Eigen::Matrix4f getFinalTransformation ()
 Get the final transformation matrix estimated by the registration method.
Eigen::Matrix4f getLastIncrementalTransformation ()
 Get the last incremental transformation matrix estimated by the registration method.
void setMaximumIterations (int nr_iterations)
 Set the maximum number of iterations the internal optimization should run for.
int getMaximumIterations ()
 Get the maximum number of iterations the internal optimization should run for, as set by the user.
void setRANSACOutlierRejectionThreshold (double inlier_threshold)
 Set the inlier distance threshold for the internal RANSAC outlier rejection loop.
double getRANSACOutlierRejectionThreshold ()
 Get the inlier distance threshold for the internal outlier rejection loop as set by the user.
void setMaxCorrespondenceDistance (double distance_threshold)
 Set the maximum distance threshold between two correspondent points in source <-> target.
double getMaxCorrespondenceDistance ()
 Get the maximum distance threshold between two correspondent points in source <-> target.
void setTransformationEpsilon (double epsilon)
 Set the transformation epsilon (maximum allowable difference between two consecutive transformations) in order for an optimization to be considered as having converged to the final solution.
double getTransformationEpsilon ()
 Get the transformation epsilon (maximum allowable difference between two consecutive transformations) as set by the user.
void setEuclideanFitnessEpsilon (double epsilon)
 Set the maximum allowed Euclidean error between two consecutive steps in the ICP loop, before the algorithm is considered to have converged.
double getEuclideanFitnessEpsilon ()
 Get the maximum allowed distance error before the algorithm will be considered to have converged, as set by the user.
void setPointRepresentation (const PointRepresentationConstPtr &point_representation)
 Provide a boost shared pointer to the PointRepresentation to be used when comparing points.
bool registerVisualizationCallback (boost::function< FunctionSignature > &visualizerCallback)
 Register the user callback function which will be called from registration thread in order to update point cloud obtained after each iteration.
double getFitnessScore (double max_range=std::numeric_limits< double >::max())
 Obtain the Euclidean fitness score (e.g., sum of squared distances from the source to the target)
double getFitnessScore (const std::vector< float > &distances_a, const std::vector< float > &distances_b)
 Obtain the Euclidean fitness score (e.g., sum of squared distances from the source to the target) from two sets of correspondence distances (distances between source and target points)
bool hasConverged ()
 Return the state of convergence after the last align run.
void align (PointCloudSource &output)
 Call the registration algorithm which estimates the transformation and returns the transformed source (input) as output.
void align (PointCloudSource &output, const Eigen::Matrix4f &guess)
 Call the registration algorithm which estimates the transformation and returns the transformed source (input) as output.
const std::string & getClassName () const
 Abstract class get name method.

Detailed Description

template<typename PointSource, typename PointTarget>
class pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >

GeneralizedIterativeClosestPoint is an ICP variant that implements the generalized iterative closest point algorithm as described by Alex Segal et al.

GeneralizedIterativeClosestPoint is an implementation of the Generalized Iterative Closest Point Algorithm

in http://www.stanford.edu/~avsegal/resources/papers/Generalized_ICP.pdf The approach is based on using anistropic cost functions to optimize the alignment after closest point assignments have been made. The original code uses GSL and ANN while in ours BFGS is replaced with a LM and ANN with FLANN.

Author:
Nizar Sallem
Radu Bogdan Rusu, Michael Dixon, Dirk Holz

Definition at line 62 of file stanford_gicp.h.


Member Typedef Documentation

typedef boost::shared_ptr< const Registration<PointSource, PointTarget> > pcl::Registration::ConstPtr [inherited]

Definition at line 70 of file registration.h.

typedef pcl::KdTree<PointTarget> pcl::Registration::KdTree [inherited]

Definition at line 72 of file registration.h.

typedef pcl::KdTree<PointTarget>::Ptr pcl::Registration::KdTreePtr [inherited]

Definition at line 73 of file registration.h.

Definition at line 83 of file registration.h.

typedef boost::shared_ptr< Registration<PointSource, PointTarget> > pcl::Registration::Ptr [inherited]

Definition at line 69 of file registration.h.

Definition at line 85 of file registration.h.

Definition at line 87 of file registration.h.

Definition at line 86 of file registration.h.


Constructor & Destructor Documentation

template<typename PointSource, typename PointTarget>
pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::GeneralizedIterativeClosestPoint ( ) [inline]

Empty constructor.

Definition at line 96 of file gicp.h.

template<typename PointSource, typename PointTarget>
pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::GeneralizedIterativeClosestPoint ( ) [inline]

Empty constructor.

Definition at line 92 of file stanford_gicp.h.


Member Function Documentation

void pcl::Registration::align ( PointCloudSource output) [inline, inherited]

Call the registration algorithm which estimates the transformation and returns the transformed source (input) as output.

Parameters:
outputthe resultant input transfomed point cloud dataset
void pcl::Registration::align ( PointCloudSource output,
const Eigen::Matrix4f &  guess 
) [inline, inherited]

Call the registration algorithm which estimates the transformation and returns the transformed source (input) as output.

Parameters:
outputthe resultant input transfomed point cloud dataset
guessthe initial gross estimation of the transformation
template<typename PointSource , typename PointTarget >
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::computeRDerivative ( const double  x[],
const Eigen::Matrix3d &  R,
double  g[] 
)

Computes rotation matrix derivative.

rotation matrix is obtainded from rotation angles x[3], x[4] and x[5]

Returns:
d/d_rx, d/d_ry and d/d_rz respectively in g[3], g[4] and g[5] param x array representing 3D transformation param R rotation matrix param g gradient vector

Definition at line 121 of file gicp.hpp.

template<typename PointSource , typename PointTarget >
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::estimateRigidTransformationLM ( const PointCloudSource cloud_src,
const PointCloudTarget cloud_tgt,
Eigen::Matrix4f &  transformation_matrix 
)

Estimate a rigid rotation transformation between a source and a target point cloud using an iterative non-linear Levenberg-Marquardt approach.

Parameters:
cloud_srcthe source point cloud dataset
cloud_tgtthe target point cloud dataset
transformation_matrixthe resultant transformation matrix

Definition at line 179 of file gicp.hpp.

template<typename PointSource , typename PointTarget >
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::estimateRigidTransformationLM ( const PointCloudSource cloud_src,
const std::vector< int > &  indices_src,
const PointCloudTarget cloud_tgt,
const std::vector< int > &  indices_tgt,
Eigen::Matrix4f &  transformation_matrix 
)

Estimate a rigid rotation transformation between a source and a target point cloud using an iterative non-linear Levenberg-Marquardt approach.

Parameters:
cloud_srcthe source point cloud dataset
indices_srcthe vector of indices describing the points of interest in cloud_src
cloud_tgtthe target point cloud dataset
indices_tgtthe vector of indices describing the correspondences of the interst points from indices_src
transformation_matrixthe resultant transformation matrix

Definition at line 240 of file gicp.hpp.

const std::string& pcl::Registration::getClassName ( ) const [inline, inherited]

Abstract class get name method.

Definition at line 263 of file registration.h.

template<typename PointSource, typename PointTarget>
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::getCorrespondenceRandomness ( ) [inline]

Get the number of neighbors used when computing covariances as set by the user.

Definition at line 213 of file gicp.h.

double pcl::Registration::getEuclideanFitnessEpsilon ( ) [inline, inherited]

Get the maximum allowed distance error before the algorithm will be considered to have converged, as set by the user.

See setEuclideanFitnessEpsilon

Definition at line 200 of file registration.h.

Eigen::Matrix4f pcl::Registration::getFinalTransformation ( ) [inline, inherited]

Get the final transformation matrix estimated by the registration method.

Definition at line 126 of file registration.h.

double pcl::Registration::getFitnessScore ( double  max_range = std::numeric_limits<double>::max ()) [inline, inherited]

Obtain the Euclidean fitness score (e.g., sum of squared distances from the source to the target)

Parameters:
max_rangemaximum allowable distance between a point and its correspondence in the target (default: double::max)
double pcl::Registration::getFitnessScore ( const std::vector< float > &  distances_a,
const std::vector< float > &  distances_b 
) [inline, inherited]

Obtain the Euclidean fitness score (e.g., sum of squared distances from the source to the target) from two sets of correspondence distances (distances between source and target points)

Parameters:
[in]distances_athe first set of distances between correspondences
[in]distances_bthe second set of distances between correspondences
PointCloudTargetConstPtr const pcl::Registration::getInputTarget ( ) [inline, inherited]

Get a pointer to the input point cloud dataset target.

Definition at line 122 of file registration.h.

Eigen::Matrix4f pcl::Registration::getLastIncrementalTransformation ( ) [inline, inherited]

Get the last incremental transformation matrix estimated by the registration method.

Definition at line 130 of file registration.h.

double pcl::Registration::getMaxCorrespondenceDistance ( ) [inline, inherited]

Get the maximum distance threshold between two correspondent points in source <-> target.

If the distance is larger than this threshold, the points will be ignored in the alignment process.

Definition at line 168 of file registration.h.

int pcl::Registration::getMaximumIterations ( ) [inline, inherited]

Get the maximum number of iterations the internal optimization should run for, as set by the user.

Definition at line 140 of file registration.h.

double pcl::Registration::getRANSACOutlierRejectionThreshold ( ) [inline, inherited]

Get the inlier distance threshold for the internal outlier rejection loop as set by the user.

Definition at line 154 of file registration.h.

template<typename PointSource, typename PointTarget>
double pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::getRotationEpsilon ( ) [inline]

Get the rotation epsilon (maximum allowable difference between two consecutive rotations) as set by the user.

Definition at line 198 of file gicp.h.

double pcl::Registration::getTransformationEpsilon ( ) [inline, inherited]

Get the transformation epsilon (maximum allowable difference between two consecutive transformations) as set by the user.

Definition at line 183 of file registration.h.

bool pcl::Registration::hasConverged ( ) [inline, inherited]

Return the state of convergence after the last align run.

Definition at line 244 of file registration.h.

template<typename PointSource, typename PointTarget>
const Eigen::Matrix3d& pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::mahalanobis ( size_t  index) const [inline]
Returns:
Mahalanobis distance matrix for the given point index

Definition at line 170 of file gicp.h.

bool pcl::Registration::registerVisualizationCallback ( boost::function< FunctionSignature > &  visualizerCallback) [inline, inherited]

Register the user callback function which will be called from registration thread in order to update point cloud obtained after each iteration.

Parameters:
refferenceof the user callback function

Definition at line 216 of file registration.h.

template<typename PointSource, typename PointTarget>
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::setCorrespondenceRandomness ( int  k) [inline]

Set the number of neighbors used when selecting a point neighbourhood to compute covariances.

A higher value will bring more accurate covariance matrix but will make covariances computation slower.

Parameters:
kthe number of neighbors to use when computing covariances

Definition at line 207 of file gicp.h.

void pcl::Registration::setEuclideanFitnessEpsilon ( double  epsilon) [inline, inherited]

Set the maximum allowed Euclidean error between two consecutive steps in the ICP loop, before the algorithm is considered to have converged.

The error is estimated as the sum of the differences between correspondences in an Euclidean sense, divided by the number of correspondences.

Parameters:
epsilonthe maximum allowed distance error before the algorithm will be considered to have converged

Definition at line 194 of file registration.h.

template<typename PointSource, typename PointTarget>
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::setInputCloud ( const PointCloudSourceConstPtr &  cloud) [inline]

Provide a pointer to the input dataset.

Parameters:
cloudthe const boost shared pointer to a PointCloud message

Definition at line 116 of file gicp.h.

template<typename PointSource, typename PointTarget>
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::setInputTarget ( const PointCloudTargetConstPtr &  target) [inline, virtual]

Provide a pointer to the input target (e.g., the point cloud that we want to align the input source to)

Parameters:
cloudthe input point cloud target

Reimplemented from pcl::Registration< PointSource, PointTarget >.

Definition at line 137 of file gicp.h.

void pcl::Registration::setMaxCorrespondenceDistance ( double  distance_threshold) [inline, inherited]

Set the maximum distance threshold between two correspondent points in source <-> target.

If the distance is larger than this threshold, the points will be ignored in the alignment process.

Parameters:
distance_thresholdthe maximum distance threshold between a point and its nearest neighbor correspondent in order to be considered in the alignment process

Definition at line 162 of file registration.h.

template<typename PointSource, typename PointTarget>
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::setMaxDistance ( double  max_distance) [inline]

Definition at line 98 of file stanford_gicp.h.

void pcl::Registration::setMaximumIterations ( int  nr_iterations) [inline, inherited]

Set the maximum number of iterations the internal optimization should run for.

Parameters:
nr_iterationsthe maximum number of iterations the internal optimization should run for

Definition at line 136 of file registration.h.

void pcl::Registration::setPointRepresentation ( const PointRepresentationConstPtr point_representation) [inline, inherited]

Provide a boost shared pointer to the PointRepresentation to be used when comparing points.

Parameters:
point_representationthe PointRepresentation to be used by the k-D tree

Definition at line 206 of file registration.h.

void pcl::Registration::setRANSACOutlierRejectionThreshold ( double  inlier_threshold) [inline, inherited]

Set the inlier distance threshold for the internal RANSAC outlier rejection loop.

The method considers a point to be an inlier, if the distance between the target data index and the transformed source index is smaller than the given inlier distance threshold. The value is set by default to 0.05m.

Parameters:
inlier_thresholdthe inlier distance threshold for the internal RANSAC outlier rejection loop

Definition at line 150 of file registration.h.

template<typename PointSource, typename PointTarget>
void pcl::GeneralizedIterativeClosestPoint< PointSource, PointTarget >::setRotationEpsilon ( double  epsilon) [inline]

Set the rotation epsilon (maximum allowable difference between two consecutive rotations) in order for an optimization to be considered as having converged to the final solution.

Parameters:
epsilonthe rotation epsilon

Definition at line 192 of file gicp.h.

void pcl::Registration::setTransformationEpsilon ( double  epsilon) [inline, inherited]

Set the transformation epsilon (maximum allowable difference between two consecutive transformations) in order for an optimization to be considered as having converged to the final solution.

Parameters:
epsilonthe transformation epsilon in order for an optimization to be considered as having converged to the final solution.

Definition at line 177 of file registration.h.

void pcl::Registration::setTransformationEstimation ( const TransformationEstimationPtr te) [inline, inherited]

Definition at line 112 of file registration.h.


The documentation for this class was generated from the following files:
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Defines