Four Points Congruent Sets (4PCS) registration algorithm (Dror Aiger, Niloy J. Mitra, Daniel Cohen-Or) More...

#include <RegistrationTools.h>

Inheritance diagram for CCCoreLib::FPCSRegistrationTools:

Collaboration diagram for CCCoreLib::FPCSRegistrationTools:

Classes
struct	Base
	FCPS base. More...

Static Public Member Functions
static bool	RegisterClouds (GenericIndexedCloud modelCloud, GenericIndexedCloud dataCloud, ScaledTransformation &transform, ScalarType delta, ScalarType beta, PointCoordinateType overlap, unsigned nbBases, unsigned nbTries, GenericProgressCallback *progressCb=nullptr, unsigned nbMaxCandidates=0)
	Registers two point clouds. More...

Static Public Member Functions inherited from CCCoreLib::RegistrationTools
static void	FilterTransformation (const ScaledTransformation &inTrans, int transformationFilters, ScaledTransformation &outTrans)
	'Filters' a transformation by constraining it along certain rotation axes and translation directions More...

Static Protected Member Functions
static bool	FindBase (GenericIndexedCloud *cloud, PointCoordinateType overlap, unsigned nbTries, Base &base)
	Randomly finds a 4 points base in a cloud. More...

static int	FindCongruentBases (KDTree tree, ScalarType delta, const CCVector3 base[4], std::vector< Base > &results)

static unsigned	ComputeRegistrationScore (KDTree modelTree, GenericIndexedCloud dataCloud, ScalarType delta, const ScaledTransformation &dataToModel)
	Registration score computation function. More...

static bool	LinesIntersections (const CCVector3 &p0, const CCVector3 &p1, const CCVector3 &p2, const CCVector3 &p3, CCVector3 &inter, PointCoordinateType &lambda, PointCoordinateType &mu)
	Find the 3D pseudo intersection between two lines. More...

static bool	FilterCandidates (GenericIndexedCloud modelCloud, GenericIndexedCloud dataCloud, Base &reference, std::vector< Base > &candidates, unsigned nbMaxCandidates, std::vector< ScaledTransformation > &transforms)

Static Protected Member Functions inherited from CCCoreLib::RegistrationTools
static bool	RegistrationProcedure (GenericCloud P, GenericCloud X, ScaledTransformation &trans, bool adjustScale=false, ScalarField *coupleWeights=nullptr, PointCoordinateType aPrioriScale=1.0f)
	ICP Registration procedure with optional scale estimation. More...

Additional Inherited Members
Public Types inherited from CCCoreLib::RegistrationTools
enum	TRANSFORMATION_FILTERS { SKIP_NONE = 0 , SKIP_RXY = 1 , SKIP_RYZ = 2 , SKIP_RXZ = 4 , SKIP_ROTATION = 7 , SKIP_TX = 8 , SKIP_TY = 16 , SKIP_TZ = 32 , SKIP_TRANSLATION = 56 }
	Transformation constraints.

using	ScaledTransformation = PointProjectionTools::Transformation
	Shortcut to PointProjectionTools::ScaledTransformation.

Detailed Description

Four Points Congruent Sets (4PCS) registration algorithm (Dror Aiger, Niloy J. Mitra, Daniel Cohen-Or)

Member Function Documentation

◆ ComputeRegistrationScore()

unsigned FPCSRegistrationTools::ComputeRegistrationScore	(	KDTree *	modelTree,
		GenericIndexedCloud *	dataCloud,
		ScalarType	delta,
		const ScaledTransformation &	dataToModel
	)

staticprotected

Registration score computation function.

!

Parameters

modelTree	KD-tree containing the model point cloud
dataCloud	data point cloud
dataToModel	transformation that, applied to data points, register model and data clouds
delta	tolerance above which data points are not counted (if a point is less than delta-apart from the model cloud, then it is counted)

Returns: the number of data points which are distance-apart from the model cloud

◆ FilterCandidates()

bool FPCSRegistrationTools::FilterCandidates	(	GenericIndexedCloud *	modelCloud,
		GenericIndexedCloud *	dataCloud,
		Base &	reference,
		std::vector< Base > &	candidates,
		unsigned	nbMaxCandidates,
		std::vector< ScaledTransformation > &	transforms
	)

staticprotected

!function to keep only the N best candidates bases (by comparison with the reference base invariants) Let B1 and B2 be 2 candidates, R be the reference, B1 and B2 aligned with R. B1 is better than B2 if the distance between B1 and R points is smaller than distance between B2 and R points. This function also computes and store the rigid transforms that align candidates with reference base

Parameters

modelCloud	the model point cloud to work on
dataCloud	the data point cloud to work on
reference	reference base
candidates	array of candidates bases. At the end of the function, this array contains the nbMaxBases best candidates only
nbMaxCandidates	maximal number of candidates allowed (if 0, number of candidates is not bounded)
transforms	array of rigid transforms that align candidates bases with the reference base

Returns: false if something went wrong

◆ FindBase()

bool FPCSRegistrationTools::FindBase	(	GenericIndexedCloud *	cloud,
		PointCoordinateType	overlap,
		unsigned	nbTries,
		Base &	base
	)

staticprotected

Randomly finds a 4 points base in a cloud.

Parameters

cloud	the point cloud in which we want to find a base
overlap	estimation of the overlap rate
nbTries	the maximum number of tries to find a base
base	the resulting base

Returns: false: failure ; true: success

◆ FindCongruentBases()

int FPCSRegistrationTools::FindCongruentBases	(	KDTree *	tree,
		ScalarType	delta,
		const CCVector3 *	base[4],
		std::vector< Base > &	results
	)

staticprotected

Find bases which are congruent to a specified 4 points base

Parameters

tree	the KD-tree build from data cloud
delta	used for the tolerance when searching for congruent bases
base	the reference base made of 4 points
results	the resulting bases

Returns: the number of bases found (number of element in the results array) or -1 is a problem occurred

◆ LinesIntersections()

bool FPCSRegistrationTools::LinesIntersections	(	const CCVector3 &	p0,
		const CCVector3 &	p1,
		const CCVector3 &	p2,
		const CCVector3 &	p3,
		CCVector3 &	inter,
		PointCoordinateType &	lambda,
		PointCoordinateType &	mu
	)

staticprotected

Find the 3D pseudo intersection between two lines.

This function finds the 3D point which is the nearest from the both lines (when this point is unique, i.e. when the lines are not parallel)

Parameters

p0	first of the two distinct points defining the first line (lying on the line)
p1	second of the two distinct points defining the first line (lying on the line)
p2	first of the two distinct points defining the second line
p3	first of the two distinct points defining the second line
inter	[out] is the computed intersection (function output)
lambda	[out] coeff such that p0+lambda(p1-p0) is the point of [p0, p1] which is the nearest from [p2, p3]
mu	[out] coeff such that p2+mu(p3-p2) is the point of [p2, p3] which is the nearest from [p0, p1]

Returns: false: no intersection was found (lines may be parallel); true: inter is the pseudo intersection

◆ RegisterClouds()

bool FPCSRegistrationTools::RegisterClouds	(	GenericIndexedCloud *	modelCloud,
		GenericIndexedCloud *	dataCloud,
		ScaledTransformation &	transform,
		ScalarType	delta,
		ScalarType	beta,
		PointCoordinateType	overlap,
		unsigned	nbBases,
		unsigned	nbTries,
		GenericProgressCallback *	progressCb = `nullptr`,
		unsigned	nbMaxCandidates = `0`
	)

static

Registers two point clouds.

Implements the 4 Points Congruent Sets Algorithm (Dror Aiger, Niloy J. Mitra, Daniel Cohen-Or

Parameters

modelCloud	the reference cloud (won't move)
dataCloud	the cloud to register (will move)
transform	the resulting transformation (output)
delta	maximal distance to the reference cloud for the data points to be considered as registered
beta	is used for bases selection (error tolerance)
overlap	estimation of the two clouds overlap rate
nbBases	number of iteration for the algorithm
nbTries	number of tries to find a base in the reference cloud
progressCb	the client application can get some notification of the process progress through this callback mechanism (see GenericProgressCallback)
nbMaxCandidates	if>0, maximal number of candidate bases allowed for each step. Otherwise the number of candidates is not bounded

Returns: false: failure ; true: success.

The documentation for this class was generated from the following files:

RegistrationTools.h
RegistrationTools.cpp

Classes

Static Public Member Functions

Static Protected Member Functions

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ ComputeRegistrationScore()

◆ FilterCandidates()

◆ FindBase()

◆ FindCongruentBases()

◆ LinesIntersections()

◆ RegisterClouds()