This class implements several J-linkage functions for pose/homography determination. More...

Inheritance diagram for Ocean::Geometry::JLinkage:

Static Public Member Functions
static bool	homographyMatrices (const ImagePoint leftImagePoints, const ImagePoint rightImagePoints, const size_t correspondences, const unsigned int width, const unsigned int height, SquareMatrices3 &homographies, const unsigned int testCandidates, const ImagePoints &leftPointForInitialModels, const Scalar squarePixelErrorAssignmentThreshold, std::vector< IndexSet32 > usedIndicesPerHomography=nullptr, bool refineHomographies=true, bool approximatedNeighborSearch=true, Ocean::RandomGenerator randomGenerator=nullptr)
	Calculates multiple homographies between two images transforming the projected planar object points between the two images using J-linkage The resulting homographies transform its image points defined in the left image to image points defined in the right image (rightPoint_i = H_i * leftPoint_i). More...

static bool	homographyMatrices (const ImagePoint leftImagePoints, const ImagePoint rightImagePoints, const size_t correspondences, const unsigned int width, const unsigned int height, SquareMatrices3 &homographies, const unsigned int testCandidates, const Indices32 &leftPointIndicesForInitialModels, const Scalar squarePixelErrorAssignmentThreshold, std::vector< IndexSet32 > usedIndicesPerHomography=nullptr, bool refineHomographies=true, bool approximatedNeighborSearch=true, Ocean::RandomGenerator randomGenerator=nullptr)
	Calculates multiple homographies between two images transforming the projected planar object points between the two images using J-linkage The resulting homographies transform its image points defined in the left image to image points defined in the right image (rightPoint_i = H_i * leftPoint_i). More...

static bool	fitLines (const ImagePoint imagePoints, const size_t pointCount, const unsigned int width, const unsigned int height, Lines2 &lines, const unsigned int testCandidates, const ImagePoints &pointForInitialModels, const Scalar pixelErrorAssignmentThreshold, std::vector< IndexSet32 > usedIndicesPerHomography=nullptr, bool approximatedNeighborSearch=true)
	Multiple line detector using J-linkage. More...

Protected Types
typedef std::pair< Scalar, Index32 >	DistancePair
	Definition of a pair combining a point index with the distance to a reference. More...

Static Protected Member Functions
static std::vector< Indices32 >	buildingMinimalSampleSet (const ImagePoint imagePoints, const size_t pointCount, const ImagePoints &pointForInitialModels, const unsigned int testCandidates, const SpatialDistribution::DistributionArray distributionImagePoints=nullptr)
	Generates minimal sample set for J-/T-Linkage. More...

static SquareMatrices3	buildingMinimalSampleSetHomography (const ImagePoint leftImagePoints, const ImagePoint rightImagePoints, const size_t correspondences, const ImagePoints &leftPointForInitialModels, const unsigned int testCandidates, const SpatialDistribution::DistributionArray distributionImagePoints=nullptr, RandomGenerator randomRansac=nullptr)
	Generates a homography per minimal sample set (for J-/T-Linkage) More...

static Lines2	buildingMinimalSampleSetLine (const ImagePoint imagePoints, const size_t pointCount, const ImagePoints &pointForInitialModels, const unsigned int testCandidates, const SpatialDistribution::DistributionArray distributionImagePoints=nullptr)
	Generates a line model per minimal sample set (for J-/T-Linkage) More...

static Scalar	jaccardDistance (const IndexSet32 &setA, const IndexSet32 &setB)
	Calculates the jaccard distance d(A, B) = ( \|union(A, B)\| - \|intersection(A, B)\| ) / \|union(A, B)\|. More...

static bool	distancePairSortAscending (const DistancePair &firstPair, const DistancePair &seccondPair)
	Sorts pairs of indices and their corresponding distances in ascending order regarding the distance values. More...

Detailed Description

This class implements several J-linkage functions for pose/homography determination.

J-linkage aims to fit multiple models from data given The J-linkage algorithm is an agglomerative clustering that proceeds by linking elements with Jaccard distance smaller than 1 and stop as soon as there are no such elements left.

Member Typedef Documentation

◆ DistancePair

typedef std::pair<Scalar, Index32> Ocean::Geometry::JLinkage::DistancePair

protected

Definition of a pair combining a point index with the distance to a reference.

Member Function Documentation

◆ buildingMinimalSampleSet()

static std::vector<Indices32> Ocean::Geometry::JLinkage::buildingMinimalSampleSet	(	const ImagePoint *	imagePoints,
		const size_t	pointCount,
		const ImagePoints &	pointForInitialModels,
		const unsigned int	testCandidates,
		const SpatialDistribution::DistributionArray *	distributionImagePoints = `nullptr`
	)

staticprotected

Generates minimal sample set for J-/T-Linkage.

Parameters

imagePoints	Image points
pointCount	Number of points
pointForInitialModels	Initial points of minimal sample sets. Note: some maybe be discarded
testCandidates	Number of candidates used for each minimal sampling set, with range [1, correspondences]
distributionImagePoints	Spatial distribution for approximate search (if nullptr, brute force methode is used)

Returns: List of minimal sample sets (represented by indices)

◆ buildingMinimalSampleSetHomography()

static SquareMatrices3 Ocean::Geometry::JLinkage::buildingMinimalSampleSetHomography	(	const ImagePoint *	leftImagePoints,
		const ImagePoint *	rightImagePoints,
		const size_t	correspondences,
		const ImagePoints &	leftPointForInitialModels,
		const unsigned int	testCandidates,
		const SpatialDistribution::DistributionArray *	distributionImagePoints = `nullptr`,
		RandomGenerator *	randomRansac = `nullptr`
	)

staticprotected

Generates a homography per minimal sample set (for J-/T-Linkage)

Parameters

leftImagePoints	Image points in the left camera, each point corresponds to one point in the right image
rightImagePoints	Image points in the right camera
correspondences	Number of points correspondences
leftPointForInitialModels	Initial points of minimal sample sets. Note: some maybe be discarded
testCandidates	Number of candidates used for each minimal sampling set, with range [4, correspondences]
distributionImagePoints	Spatial distribution for approximate search (if nullptr, brute force methode is used)
randomRansac	Random generator object to be used for creating random numbers, RANSAC is used if this is not null

Returns: Resulting homographies for each minimal sample set (its count is possibly smaller then the given image points)

◆ buildingMinimalSampleSetLine()

static Lines2 Ocean::Geometry::JLinkage::buildingMinimalSampleSetLine	(	const ImagePoint *	imagePoints,
		const size_t	pointCount,
		const ImagePoints &	pointForInitialModels,
		const unsigned int	testCandidates,
		const SpatialDistribution::DistributionArray *	distributionImagePoints = `nullptr`
	)

staticprotected

Generates a line model per minimal sample set (for J-/T-Linkage)

Parameters

imagePoints	Image points
pointCount	Number of points
pointForInitialModels	Initial points of minimal sample sets. Note: some maybe be discarded
testCandidates	Number of candidates used for each minimal sampling set, with range [2, correspondences]
distributionImagePoints	Spatial distribution for approximate search (if nullptr, brute force methode is used)

Returns: Resulting line for each minimal sample set (its count is possibly smaller then the given image points)

◆ distancePairSortAscending()

bool Ocean::Geometry::JLinkage::distancePairSortAscending	(	const DistancePair &	firstPair,
		const DistancePair &	seccondPair
	)

inlinestaticprotected

Sorts pairs of indices and their corresponding distances in ascending order regarding the distance values.

Parameters

firstPair	First pair of distance and index
seccondPair	Second pair of distance and index

Returns: True, if the distance value of the first pair is smaller than distance value of the second pair

◆ fitLines()

static bool Ocean::Geometry::JLinkage::fitLines	(	const ImagePoint *	imagePoints,
		const size_t	pointCount,
		const unsigned int	width,
		const unsigned int	height,
		Lines2 &	lines,
		const unsigned int	testCandidates,
		const ImagePoints &	pointForInitialModels,
		const Scalar	pixelErrorAssignmentThreshold,
		std::vector< IndexSet32 > *	usedIndicesPerHomography = `nullptr`,
		bool	approximatedNeighborSearch = `true`
	)

static

Multiple line detector using J-linkage.

Parameters

imagePoints	image points
pointCount	Number of points
width	The width of the image in pixel
height	The height of the image in pixel
lines	Resulting valid line model
testCandidates	Number of candidates used in for minimum sample set, with range [2, correspondences]
pointForInitialModels	Initial image points, one per each sample set
pixelErrorAssignmentThreshold	Maximal pixel error of a point-line corresspodance (0, infinity)
usedIndicesPerHomography	Optional set of indices which will receive the indices of the used image correspondences per model, if defined
approximatedNeighborSearch	Defines if speeded up spatial neighbor search is used

Returns: True, if successfully completed

◆ homographyMatrices() [1/2]

static bool Ocean::Geometry::JLinkage::homographyMatrices	(	const ImagePoint *	leftImagePoints,
		const ImagePoint *	rightImagePoints,
		const size_t	correspondences,
		const unsigned int	width,
		const unsigned int	height,
		SquareMatrices3 &	homographies,
		const unsigned int	testCandidates,
		const ImagePoints &	leftPointForInitialModels,
		const Scalar	squarePixelErrorAssignmentThreshold,
		std::vector< IndexSet32 > *	usedIndicesPerHomography = `nullptr`,
		bool	refineHomographies = `true`,
		bool	approximatedNeighborSearch = `true`,
		Ocean::RandomGenerator *	randomGenerator = `nullptr`
	)

static

Calculates multiple homographies between two images transforming the projected planar object points between the two images using J-linkage The resulting homographies transform its image points defined in the left image to image points defined in the right image (rightPoint_i = H_i * leftPoint_i).

Parameters

leftImagePoints	Image points in the left camera, each point corresponds to one point in the right image
rightImagePoints	Image points in the right camera
correspondences	Number of points correspondences
width	The width of the left image in pixel
height	The height of the left image in pixel
homographies	Resulting valid homographies for each image plane
testCandidates	Number of candidates used in for minimum sample set, with range [4, correspondences]
leftPointForInitialModels	Initial image points, one per each sample set
squarePixelErrorAssignmentThreshold	Maximal square pixel error of a valid projection of a 3D point onto the 2D image plane, with range (0, infinity)
usedIndicesPerHomography	Optional set of indices which will receive the indices of the used image correspondences per model, if defined
refineHomographies	Determines whether a not linear least square algorithm is used to increase the pose accuracies after J-linkage
approximatedNeighborSearch	Defines if speeded up spatial neighbor search is used
randomGenerator	Random number generator. If is not nullptr, RANSAC is used for homography determination within initial minimum sample set

Returns: True, if successfully completed

◆ homographyMatrices() [2/2]

bool Ocean::Geometry::JLinkage::homographyMatrices	(	const ImagePoint *	leftImagePoints,
		const ImagePoint *	rightImagePoints,
		const size_t	correspondences,
		const unsigned int	width,
		const unsigned int	height,
		SquareMatrices3 &	homographies,
		const unsigned int	testCandidates,
		const Indices32 &	leftPointIndicesForInitialModels,
		const Scalar	squarePixelErrorAssignmentThreshold,
		std::vector< IndexSet32 > *	usedIndicesPerHomography = `nullptr`,
		bool	refineHomographies = `true`,
		bool	approximatedNeighborSearch = `true`,
		Ocean::RandomGenerator *	randomGenerator = `nullptr`
	)

inlinestatic

Calculates multiple homographies between two images transforming the projected planar object points between the two images using J-linkage The resulting homographies transform its image points defined in the left image to image points defined in the right image (rightPoint_i = H_i * leftPoint_i).

Parameters

leftImagePoints	Image points in the left camera, each point corresponds to one point in the right image
rightImagePoints	Image points in the right camera
correspondences	Number of points correspondences
width	The width of the left image in pixel
height	The height of the left image in pixel
homographies	Resulting valid homographies for each image plane
testCandidates	Number of candidates used in for minimum sample set, with range [4, correspondences]
leftPointIndicesForInitialModels	Initial indices of image points, one per each sample set
squarePixelErrorAssignmentThreshold	Maximal square pixel error of a valid projection of a 3D point onto the 2D image plane, with range (0, infinity)
usedIndicesPerHomography	Optional set of indices which will receive the indices of the used image correspondences per model, if defined
refineHomographies	Determines whether a not linear least square algorithm is used to increase the pose accuracies after J-linkage
approximatedNeighborSearch	Defines if speeded up spatial neighbor search is used
randomGenerator	Random number generator. If is not nullptr, RANSAC is used for homography determination within initial minimum sample set

Returns: True, if successfully completed

◆ jaccardDistance()

Scalar Ocean::Geometry::JLinkage::jaccardDistance	(	const IndexSet32 &	setA,
		const IndexSet32 &	setB
	)

inlinestaticprotected

Calculates the jaccard distance d(A, B) = ( |union(A, B)| - |intersection(A, B)| ) / |union(A, B)|.

Parameters

setA	first set of indices to compare
setB	second set of indices to compare

Returns: Jaccard distance [0, 1]

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

JLinkage.h

Static Public Member Functions

Protected Types

Static Protected Member Functions

Detailed Description

Member Typedef Documentation

◆ DistancePair

Member Function Documentation

◆ buildingMinimalSampleSet()

◆ buildingMinimalSampleSetHomography()

◆ buildingMinimalSampleSetLine()

◆ distancePairSortAscending()

◆ fitLines()

◆ homographyMatrices() [1/2]

◆ homographyMatrices() [2/2]

◆ jaccardDistance()