DefaultConvergenceCriteria represents an instantiation of ConvergenceCriteria, and implements the following criteria for registration loop evaluation: More...

#include <pcl/registration/default_convergence_criteria.h>

Inheritance diagram for pcl::registration::DefaultConvergenceCriteria< Scalar >:

Public Types
enum	ConvergenceState { CONVERGENCE_CRITERIA_NOT_CONVERGED , CONVERGENCE_CRITERIA_ITERATIONS , CONVERGENCE_CRITERIA_TRANSFORM , CONVERGENCE_CRITERIA_ABS_MSE , CONVERGENCE_CRITERIA_REL_MSE , CONVERGENCE_CRITERIA_NO_CORRESPONDENCES , CONVERGENCE_CRITERIA_FAILURE_AFTER_MAX_ITERATIONS }
using	Ptr = shared_ptr<DefaultConvergenceCriteria<Scalar>>
using	ConstPtr = shared_ptr<const DefaultConvergenceCriteria<Scalar>>
using	Matrix4 = Eigen::Matrix<Scalar, 4, 4>
Public Types inherited from pcl::registration::ConvergenceCriteria
using	Ptr = shared_ptr<ConvergenceCriteria>
using	ConstPtr = shared_ptr<const ConvergenceCriteria>

Public Member Functions
	DefaultConvergenceCriteria (const int &iterations, const Matrix4 &transform, const pcl::Correspondences &correspondences)
	Empty constructor.
	~DefaultConvergenceCriteria () override=default
	Empty destructor.
void	setMaximumIterationsSimilarTransforms (const int nr_iterations)
	Set the maximum number of consecutive iterations that the internal rotation, translation, and MSE differences are allowed to be similar.
int	getMaximumIterationsSimilarTransforms () const
	Get the maximum number of consecutive iterations that the internal rotation, translation, and MSE differences are allowed to be similar, as set by the user.
void	setMaximumIterations (const int nr_iterations)
	Set the maximum number of iterations the internal optimization should run for.
int	getMaximumIterations () const
	Get the maximum number of iterations the internal optimization should run for, as set by the user.
void	setFailureAfterMaximumIterations (const bool failure_after_max_iter)
	Specifies if the registration fails or converges when the maximum number of iterations is reached.
bool	getFailureAfterMaximumIterations () const
	Get whether the registration will fail or converge when the maximum number of iterations is reached.
void	setRotationThreshold (const double threshold)
	Set the rotation threshold cosine angle (maximum allowable difference between two consecutive transformations) in order for an optimization to be considered as having converged to the final solution.
double	getRotationThreshold () const
	Get the rotation threshold cosine angle (maximum allowable difference between two consecutive transformations) as set by the user.
void	setTranslationThreshold (const double threshold)
	Set the translation threshold (maximum allowable difference between two consecutive transformations) in order for an optimization to be considered as having converged to the final solution.
double	getTranslationThreshold () const
	Get the rotation threshold cosine angle (maximum allowable difference between two consecutive transformations) as set by the user.
void	setRelativeMSE (const double mse_relative)
	Set the relative MSE between two consecutive sets of correspondences.
double	getRelativeMSE () const
	Get the relative MSE between two consecutive sets of correspondences.
void	setAbsoluteMSE (const double mse_absolute)
	Set the absolute MSE between two consecutive sets of correspondences.
double	getAbsoluteMSE () const
	Get the absolute MSE between two consecutive sets of correspondences.
bool	hasConverged () override
	Check if convergence has been reached.
ConvergenceState	getConvergenceState ()
	Return the convergence state after hasConverged ().
void	setConvergenceState (ConvergenceState c)
	Sets the convergence state externally (for example, when ICP does not find enough correspondences to estimate a transformation, the function is called setting the convergence state to ConvergenceState::CONVERGENCE_CRITERIA_NO_CORRESPONDENCES).
Public Member Functions inherited from pcl::registration::ConvergenceCriteria
	ConvergenceCriteria ()=default
	Empty constructor.
virtual	~ConvergenceCriteria ()=default
	Empty destructor.
	operator bool ()
	Bool operator.

Protected Member Functions
double	calculateMSE (const pcl::Correspondences &correspondences) const
	Calculate the mean squared error (MSE) of the distance for a given set of correspondences.

Protected Attributes
const int &	iterations_
	The number of iterations done by the registration loop so far.
const Matrix4 &	transformation_
	The current transformation obtained by the transformation estimation method.
const pcl::Correspondences &	correspondences_
	The current set of point correspondences between the source and the target.
double	correspondences_prev_mse_ {std::numeric_limits<double>::max()}
	The MSE for the previous set of correspondences.
double	correspondences_cur_mse_ {std::numeric_limits<double>::max()}
	The MSE for the current set of correspondences.
int	max_iterations_ {100}
	The maximum nuyyGmber of iterations that the registration loop is to be executed.
bool	failure_after_max_iter_ {false}
	Specifys if the registration fails or converges when the maximum number of iterations is reached.
double	rotation_threshold_ {0.99999}
	The rotation threshold is the relative rotation between two iterations (as angle cosine).
double	translation_threshold_ {3e-4 * 3e-4}
	The translation threshold is the relative translation between two iterations (0 if no translation).
double	mse_threshold_relative_ {0.00001}
	The relative change from the previous MSE for the current set of correspondences, e.g.
double	mse_threshold_absolute_ {1e-12}
	The absolute change from the previous MSE for the current set of correspondences.
int	iterations_similar_transforms_ {0}
	Internal counter for the number of iterations that the internal rotation, translation, and MSE differences are allowed to be similar.
int	max_iterations_similar_transforms_ {0}
	The maximum number of iterations that the internal rotation, translation, and MSE differences are allowed to be similar.
ConvergenceState	convergence_state_ {CONVERGENCE_CRITERIA_NOT_CONVERGED}
	The state of the convergence (e.g., why did the registration converge).

Detailed Description

template<typename Scalar = float>
class pcl::registration::DefaultConvergenceCriteria< Scalar >

DefaultConvergenceCriteria represents an instantiation of ConvergenceCriteria, and implements the following criteria for registration loop evaluation:

a maximum number of iterations has been reached
the transformation (R, t) cannot be further updated (the difference between current and previous is smaller than a threshold)
the Mean Squared Error (MSE) between the current set of correspondences and the previous one is smaller than some threshold (both relative and absolute tests)

Note: Convergence is considered reached if ANY of the above criteria are met.

Author: Radu B. Rusu

Definition at line 65 of file default_convergence_criteria.h.

Member Typedef Documentation

◆ ConstPtr

template<typename Scalar = float>

using pcl::registration::DefaultConvergenceCriteria< Scalar >::ConstPtr = shared_ptr<const DefaultConvergenceCriteria<Scalar>>

Definition at line 68 of file default_convergence_criteria.h.

◆ Matrix4

template<typename Scalar = float>

using pcl::registration::DefaultConvergenceCriteria< Scalar >::Matrix4 = Eigen::Matrix<Scalar, 4, 4>

Definition at line 70 of file default_convergence_criteria.h.

◆ Ptr

template<typename Scalar = float>

using pcl::registration::DefaultConvergenceCriteria< Scalar >::Ptr = shared_ptr<DefaultConvergenceCriteria<Scalar>>

Definition at line 67 of file default_convergence_criteria.h.

Member Enumeration Documentation

◆ ConvergenceState

template<typename Scalar = float>

enum pcl::registration::DefaultConvergenceCriteria::ConvergenceState

Enumerator
CONVERGENCE_CRITERIA_NOT_CONVERGED
CONVERGENCE_CRITERIA_ITERATIONS
CONVERGENCE_CRITERIA_TRANSFORM
CONVERGENCE_CRITERIA_ABS_MSE
CONVERGENCE_CRITERIA_REL_MSE
CONVERGENCE_CRITERIA_NO_CORRESPONDENCES
CONVERGENCE_CRITERIA_FAILURE_AFTER_MAX_ITERATIONS

Definition at line 72 of file default_convergence_criteria.h.

Constructor & Destructor Documentation

◆ DefaultConvergenceCriteria()

template<typename Scalar = float>

pcl::registration::DefaultConvergenceCriteria< Scalar >::DefaultConvergenceCriteria	(	const int &	iterations,
		const Matrix4 &	transform,
		const pcl::Correspondences &	correspondences )

inline

Empty constructor.

Sets:

the maximum number of iterations to 100
the rotation threshold to 0.256 degrees (0.99999)
the translation threshold to 0.0003 meters (3e-4^2)
the MSE relative / absolute thresholds to 0.001% and 1e-12

Parameters

[in]	iterations	a reference to the number of iterations the loop has ran so far
[in]	transform	a reference to the current transformation obtained by the transformation evaluation
[in]	correspondences	a reference to the current set of point correspondences between source and target

Definition at line 94 of file default_convergence_criteria.h.

References correspondences_, iterations_, and transformation_.

◆ ~DefaultConvergenceCriteria()

template<typename Scalar = float>

pcl::registration::DefaultConvergenceCriteria< Scalar >::~DefaultConvergenceCriteria ( )

overridedefault

Empty destructor.

Member Function Documentation

◆ calculateMSE()

template<typename Scalar = float>

double pcl::registration::DefaultConvergenceCriteria< Scalar >::calculateMSE ( const pcl::Correspondences & correspondences ) const

inlineprotected

Calculate the mean squared error (MSE) of the distance for a given set of correspondences.

Parameters

[in] correspondences the given set of correspondences

Definition at line 262 of file default_convergence_criteria.h.

Referenced by hasConverged().

◆ getAbsoluteMSE()

template<typename Scalar = float>

double pcl::registration::DefaultConvergenceCriteria< Scalar >::getAbsoluteMSE ( ) const

inline

Get the absolute MSE between two consecutive sets of correspondences.

Definition at line 230 of file default_convergence_criteria.h.

References mse_threshold_absolute_.

◆ getConvergenceState()

template<typename Scalar = float>

ConvergenceState pcl::registration::DefaultConvergenceCriteria< Scalar >::getConvergenceState ( )

inline

Return the convergence state after hasConverged ().

Definition at line 241 of file default_convergence_criteria.h.

References convergence_state_.

◆ getFailureAfterMaximumIterations()

template<typename Scalar = float>

bool pcl::registration::DefaultConvergenceCriteria< Scalar >::getFailureAfterMaximumIterations ( ) const

inline

Get whether the registration will fail or converge when the maximum number of iterations is reached.

Definition at line 156 of file default_convergence_criteria.h.

References failure_after_max_iter_.

◆ getMaximumIterations()

template<typename Scalar = float>

int pcl::registration::DefaultConvergenceCriteria< Scalar >::getMaximumIterations ( ) const

inline

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

Definition at line 138 of file default_convergence_criteria.h.

References max_iterations_.

◆ getMaximumIterationsSimilarTransforms()

template<typename Scalar = float>

int pcl::registration::DefaultConvergenceCriteria< Scalar >::getMaximumIterationsSimilarTransforms ( ) const

inline

Get the maximum number of consecutive iterations that the internal rotation, translation, and MSE differences are allowed to be similar, as set by the user.

Definition at line 120 of file default_convergence_criteria.h.

References max_iterations_similar_transforms_.

◆ getRelativeMSE()

template<typename Scalar = float>

double pcl::registration::DefaultConvergenceCriteria< Scalar >::getRelativeMSE ( ) const

inline

Get the relative MSE between two consecutive sets of correspondences.

Definition at line 214 of file default_convergence_criteria.h.

References mse_threshold_relative_.

◆ getRotationThreshold()

template<typename Scalar = float>

double pcl::registration::DefaultConvergenceCriteria< Scalar >::getRotationThreshold ( ) const

inline

Get the rotation threshold cosine angle (maximum allowable difference between two consecutive transformations) as set by the user.

Definition at line 177 of file default_convergence_criteria.h.

References rotation_threshold_.

◆ getTranslationThreshold()

template<typename Scalar = float>

double pcl::registration::DefaultConvergenceCriteria< Scalar >::getTranslationThreshold ( ) const

inline

Get the rotation threshold cosine angle (maximum allowable difference between two consecutive transformations) as set by the user.

Definition at line 198 of file default_convergence_criteria.h.

References translation_threshold_.

◆ hasConverged()

template<typename Scalar>

bool pcl::registration::DefaultConvergenceCriteria< Scalar >::hasConverged ( )

overridevirtual

◆ setAbsoluteMSE()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setAbsoluteMSE ( const double mse_absolute )

inline

Set the absolute MSE between two consecutive sets of correspondences.

Parameters

[in] mse_absolute the relative MSE threshold

Definition at line 223 of file default_convergence_criteria.h.

References mse_threshold_absolute_.

◆ setConvergenceState()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setConvergenceState ( ConvergenceState c )

inline

Sets the convergence state externally (for example, when ICP does not find enough correspondences to estimate a transformation, the function is called setting the convergence state to ConvergenceState::CONVERGENCE_CRITERIA_NO_CORRESPONDENCES).

Parameters

[in] c the convergence state

Definition at line 252 of file default_convergence_criteria.h.

References convergence_state_.

◆ setFailureAfterMaximumIterations()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setFailureAfterMaximumIterations ( const bool failure_after_max_iter )

inline

Specifies if the registration fails or converges when the maximum number of iterations is reached.

Parameters

[in] failure_after_max_iter If true, the registration fails. If false, the registration is assumed to have converged.

Definition at line 148 of file default_convergence_criteria.h.

References failure_after_max_iter_.

◆ setMaximumIterations()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setMaximumIterations ( const int nr_iterations )

inline

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

Parameters

[in] nr_iterations the maximum number of iterations the internal optimization should run for

Definition at line 130 of file default_convergence_criteria.h.

References max_iterations_.

◆ setMaximumIterationsSimilarTransforms()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setMaximumIterationsSimilarTransforms ( const int nr_iterations )

inline

Set the maximum number of consecutive iterations that the internal rotation, translation, and MSE differences are allowed to be similar.

Parameters

[in] nr_iterations the maximum number of iterations

Definition at line 110 of file default_convergence_criteria.h.

References max_iterations_similar_transforms_.

◆ setRelativeMSE()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setRelativeMSE ( const double mse_relative )

inline

Set the relative MSE between two consecutive sets of correspondences.

Parameters

[in] mse_relative the relative MSE threshold

Definition at line 207 of file default_convergence_criteria.h.

References mse_threshold_relative_.

◆ setRotationThreshold()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setRotationThreshold ( const double threshold )

inline

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

Parameters

[in] threshold the rotation threshold in order for an optimization to be considered as having converged to the final solution.

Definition at line 168 of file default_convergence_criteria.h.

References rotation_threshold_.

◆ setTranslationThreshold()

template<typename Scalar = float>

void pcl::registration::DefaultConvergenceCriteria< Scalar >::setTranslationThreshold ( const double threshold )

inline

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

Parameters

[in] threshold the translation threshold in order for an optimization to be considered as having converged to the final solution.

Definition at line 189 of file default_convergence_criteria.h.

References translation_threshold_.