opentl::math::CV Class Reference

List of all members.

Classes
class	RansacSubSetIndexComparator
	RANSAC Set Comperator function for point subsets.
Public Types
Static Public Member Functions
static void	estimate2dAffinity (math::Matrix &P1, math::Matrix &P2, math::Transform &T)
	estimates the 2D Affinity A out of the point correspondences s.t. P2 = A*P1
static void	findExtrinsicCameraParams (const math::Matrix &bodyPoints, const math::Matrix &screenPoints, const math::SquareMatrix &intrinsicMat, math::Vector3 &rotVec, math::Vector3 &traslVec, math::Vector4 *distCoeff=NULL)
	Compute Euclidean pose parameters (extrinsic camera parameters) from 3D/2D point correspondences of a calibrated camera.
static int	gpDLT (math::Transform &Trec, math::Matrix &X, math::Matrix &x, math::Transform &Tbar, math::SquareMatrix &K, math::Matrix *Q=NULL, math::Vector *p0=NULL, math::Matrix *P=NULL)
	computes P with the generalized projective DLT algorithm
static int	meanShiftClustering (math::Matrix &X, std::vector< math::Vector * > &clusters, std::vector< int > &cluster_ids, double bandWidth, MSKernelType kernelType=math::CV::GAUSS, double minNorm=0.0001, int maxIter=100, math::Vector *weights=NULL, bool bPruneNonMaxima=false)
	Mean-shift clustering algorithm.
static void	meanShiftOptimizer (math::Matrix &X, math::Vector &x0, math::Vector &x, double bandWidth, MSKernelType kernelType=math::CV::GAUSS, double minNorm=0.0001, int maxIter=100, math::Vector *weights=NULL, math::Matrix *traj=NULL)
	Internal Mean-shift optimizer.
static double	ransac2dLine (std::vector< math::Vector2 * > &inputData, std::vector< int > &idxListOfBestSamples, double delta, double desiredProbabilityForNoOutliers, double maximalOutlierPercentage, math::Vector4 &bestLineParams)
	RANdom SAmple Consensus (RANSAC) algorithm to remove outliers frome detected line segments.
static double	triangDLT (math::Vector4 *X, math::Vector *z, math::Matrix *Projmats)
	Calibrated multi-camera point triangulation based on the DLT algorithm.

---

Member Enumeration Documentation

enum opentl::math::CV::MSKernelType

Type of kernel for mean-shift: Linear, Epanechnikov, Gaussian

Enumerator:

LINEAR
EPANECH
GAUSS

---

Member Function Documentation

static void opentl::math::CV::estimate2dAffinity	(	math::Matrix &	P1,
		math::Matrix &	P2,
		math::Transform &	T
	)			[static]

estimates the 2D Affinity A out of the point correspondences s.t. P2 = A*P1

Parameters:

	P1	homogeneous 2d point correspondences part1 (3xn)
	P2	homogeneous 2d point correspondences part2 (3xn)
	T	the estimated Transformation

static void opentl::math::CV::findExtrinsicCameraParams	(	const math::Matrix &	bodyPoints,
		const math::Matrix &	screenPoints,
		const math::SquareMatrix &	intrinsicMat,
		math::Vector3 &	rotVec,
		math::Vector3 &	traslVec,
		math::Vector4 *	distCoeff = NULL
	)			[static]

Compute Euclidean pose parameters (extrinsic camera parameters) from 3D/2D point correspondences of a calibrated camera.

Parameters:

	bodyPoints	The array of object points, 3xN, where N is the number of points in the view
	screenPoints	The array of corresponding image points, 2xN or Nx2, where N is the number of points in the view
	intrinsicMat	The camera matrix (A) [fx 0 cx; 0 fy cy; 0 0 1]
	rotVec	The output 3x1 or 1x3 rotation vector (representation of a rotation matrix through the Rodrigues' formula).
	traslVec	The output 3x1 or 1x3 translation vector
	distCoeff	The vector of distortion coefficients, 4x1 or 1x4 [k1, k2, p1, p2]. If it is NULL, all distortion coefficients are considered 0's

static int opentl::math::CV::gpDLT	(	math::Transform &	Trec,
		math::Matrix &	X,
		math::Matrix &	x,
		math::Transform &	Tbar,
		math::SquareMatrix &	K,
		math::Matrix *	Q = NULL,
		math::Vector *	p0 = NULL,
		math::Matrix *	P = NULL
	)			[static]

computes P with the generalized projective DLT algorithm

Parameters:

	Trec	return reconstructed Transformation
	X	stacked homogeneous model points (4xN)
	x	stacked corresponding homogeneous screen points (3xN)
	Tbar	constant reference T matrix
	K	intrinsic 3x3 camera matrix
	Q	constraint matrix => used if given
	p0	constraint vector => used if given
	P	if matrix is given, the 3x4 projective Matrix is returned within

Returns:

0 if everything's ok, -1 if an error occured

static int opentl::math::CV::meanShiftClustering	(	math::Matrix &	X,
		std::vector< math::Vector * > &	clusters,
		std::vector< int > &	cluster_ids,
		double	bandWidth,
		MSKernelType	kernelType = math::CV::GAUSS,
		double	minNorm = 0.0001,
		int	maxIter = 100,
		math::Vector *	weights = NULL,
		bool	bPruneNonMaxima = false
	)			[static]

Mean-shift clustering algorithm.

Parameters:

	X	Input data set (d x n), where d is the vector dimension and n the number of points
	clusters	Output cluster centers
	cluster_ids	Output cluster labels for each data point
	(kernelType,bandWidth,minNorm,maxIter)	Input parameters for the mean-shift optimizer
	weights	Weights for each sample kernel
	bPruneNonMaxima	Flag: if non-maxima such as saddle points have to be removed (method: small perturbation of cluster centers, re-run mean-shift and check for convergence to the same center)

static void opentl::math::CV::meanShiftOptimizer	(	math::Matrix &	X,
		math::Vector &	x0,
		math::Vector &	x,
		double	bandWidth,
		MSKernelType	kernelType = math::CV::GAUSS,
		double	minNorm = 0.0001,
		int	maxIter = 100,
		math::Vector *	weights = NULL,
		math::Matrix *	traj = NULL
	)			[static]

Internal Mean-shift optimizer.

Parameters:

	X	Input data set (dim x N)
	x0	Input start point (dim x 1)
	x	Output estimated mode
	kernelType	Input type of kernel (linear, Epanechnikov, Gauss)
	bandWidth	Kernel width (h)
	minNorm	Convergence criterion (minimum step norm)
	maxIter	maximum number of iteration
	traj	= (optional) output trajectory during optimization

static double opentl::math::CV::ransac2dLine	(	std::vector< math::Vector2 * > &	inputData,
		std::vector< int > &	idxListOfBestSamples,
		double	delta,
		double	desiredProbabilityForNoOutliers,
		double	maximalOutlierPercentage,
		math::Vector4 &	bestLineParams
	)			[static]

RANdom SAmple Consensus (RANSAC) algorithm to remove outliers frome detected line segments.

Parameters:

	inputData	Vector of 2d points found possibly belonging to an edge
	idxListOfBestSamples	Result of RANSAC: list of best point indices found in the input data
	delta	Line thickness, as threshold for inliers, e.g. 0.5
	desiredProbabilityForNoOutliers	The probability that at least one of the selected subsets doesn't contain an outlier, e.g. 0.95 (has influence on RANSAC tries)
	maximalOutlierPercentage	The maximal expected percentage of outliers, e.g. 0.7 ((has influence on RANSAC tries)
	bestLineParams	Line parameters of best line: vec[0] = nx, normal of line vec[1] = ny, normal of line vec[2] = x-coordinate of sample point vec[3] = y-coordinate of sample point

Returns:

Returns the percentage of data used in the least squares estimate.

static double opentl::math::CV::triangDLT	(	math::Vector4 *	X,
		math::Vector *	z,
		math::Matrix *	Projmats
	)			[static]

Calibrated multi-camera point triangulation based on the DLT algorithm.

Parameters:

	X	output 3D point in world coordinates
	z	input 2D measurements (x,y) stacked in a vector
	Projmats	camera (3x4) projection matrices, stacked together