opentl::core::cvdata::T_MEAS_FEAT Class Reference

Structure holding a single feature-level measurement. A single measurement is related to: 1 target, 1 modality, 1 camera, 1 thread. More...

List of all members.

Public Types

Public Member Functions

std::size_t getFeatElementDim () const

Get size/length of a single feature element.

std::size_t getJacSize () const

Get the column-size of Jacobian matrices.

T_MEAS_FEAT_TYPE getMeasType () const

Return type of measurement.

std::size_t getNMaxAssoc () const

Get number of maximum association hypotheses (per feature).

std::size_t getNMaxMeas () const

Get number of allocated features.

std::size_t getNMeas () const

Get number of available features.

void resizeAll (T_MEAS_FEAT_TYPE measType, std::size_t nMaxMeas_, std::size_t featElementDim_, std::size_t jacSize_=0, std::size_t nMaxAssoc_=1)

Resizes all vectors within structure and sets internal variables accordingly.

void setNMeas (std::size_t nMeas_)

Set number of available features (it must be less or equal than the number of allocated features).

T_MEAS_FEAT (std::size_t nMaxMeas_=1, std::size_t featElementDim_=0, std::size_t jacSize_=0, std::size_t nMaxAssoc_=1, double alpha_=0.0, T_MEAS_FEAT_TYPE measType=VECTOR_RES_SINGLE_HYPO)

Constructor.

Public Attributes

double alpha

Probability of missing detection (0 associated measurements) for each feature.

std::vector< std::vector
< double > > E_mHypo

E_mHypo = vector of measurement residuals (or innovations) associated to each pair (h,z), with possibly multiple hypotheses NOTE: The actual number of valid residuals is given by the respective entry in nAssociations (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: local feature index
inner (2nd) [0,nAssociations[f]]: association hypotheses (NOTE: the actual number may be less than nMaxAssoc).

std::vector< double > E_sHypo

E_sHypo = vector of image residuals associated to each h, with either 0 or 1 association hypothesis. NOTE: The respective entry in nAssociations decides for the validity of this measurement (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: encodes the set of data in a single vector (in case of missing detection, the respective data are invalid).

std::vector< double > h

h = vector of expected image features (obtained for example by re-projection from object space). Dimensions:

outer (1st) [0,nMeas*featElemDim]: local feature index

std::vector< double > invCov

invCov = inverse of measurement covariances, for each feature. ATTENTION: invCov encodes a sequence of (nMeas) square matrices, each one of size (featElementDim * featElementDim), into a single vector. Therefore, the total dimension of invCov is (nMeas*featElementDim*featElementDim).

std::vector< int > nAssociations

nAssociations = vector of integers, giving the actual number of associated hypotheses per feature, between [0,nMaxAssoc].

std::vector< std::vector
< double > > transpJac

transpJac = transposed Jacobians of h. See the jacSize parameter in resizeAll(). Dimensions:

outer (1st) [0,jacSize]: state variable index
inner (2nd) [0,nMeas*featElemDim]: local feature index (in case of missing detection, data in this vector are invalid).

std::vector< std::vector
< double > > z_mHypo

z_mHypo = vector of image features associated to each h, with possibly multiple hypotheses NOTE: The actual number of valid associations is given by the respective entry in nAssociations (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: local feature index
inner (2nd) [0,nAssociations[f]]: association hypotheses (NOTE: the actual number may be less than nMaxAssoc).

std::vector< double > z_sHypo

z_sHypo = vector of image features associated to each h, with either 0 or 1 association hypothesis. NOTE: The respective entry in nAssociations decides for the validity of this measurement (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: encodes the set of data in a single vector (in case of missing detection, the respective data are invalid).

Detailed Description

Structure holding a single feature-level measurement. A single measurement is related to: 1 target, 1 modality, 1 camera, 1 thread.

Member Enumeration Documentation

enum opentl::core::cvdata::T_MEAS_FEAT::T_MEAS_FEAT_TYPE

Possible measurement data types.

Enumerator:

SCALAR_RES_SINGLE_HYPO	A unique, scalar residual for the whole measurement (but possibly vector-valued z,h) with a single association hypothesis.
VECTOR_RES_SINGLE_HYPO	Vector residuals (one vector per feature, like z-h), with a single association hypothesis (default).
VECTOR_RES_MULTI_HYPO	Vector residuals (one vector per feature, like z-h), with possibly multiple association hypotheses.

Constructor & Destructor Documentation

opentl::core::cvdata::T_MEAS_FEAT::T_MEAS_FEAT	(	std::size_t	nMaxMeas_ = `1`,
		std::size_t	featElementDim_ = `0`,
		std::size_t	jacSize_ = `0`,
		std::size_t	nMaxAssoc_ = `1`,
		double	alpha_ = `0.0`,
		T_MEAS_FEAT_TYPE	measType = `VECTOR_RES_SINGLE_HYPO`
	)			`[inline]`

Constructor.

Parameters:

	nMaxMeas_	Maximum number of features (in order to pre-allocate memory). The actual number of stored measurements is given by nMeas.
	nMaxAssoc_	Maximum number of association hypotheses (in case of multiple hypotheses, in order to pre-allocate the vectors).
	featElementDim_	Length of one feature element.
	jacSize_	Column-size of the Jacobian matrix (usually the state, or pose dimension). The row-size is featElementDim.
	alpha_	Probability of missing detection (default = 0).
	measType	Type of measurement: scalar or vector residual, with single or multiple association hypotheses.

Member Function Documentation

std::size_t opentl::core::cvdata::T_MEAS_FEAT::getFeatElementDim ( ) const [inline]

Get size/length of a single feature element.

std::size_t opentl::core::cvdata::T_MEAS_FEAT::getJacSize ( ) const [inline]

Get the column-size of Jacobian matrices.

T_MEAS_FEAT_TYPE opentl::core::cvdata::T_MEAS_FEAT::getMeasType ( ) const [inline]

Return type of measurement.

std::size_t opentl::core::cvdata::T_MEAS_FEAT::getNMaxAssoc ( ) const [inline]

Get number of maximum association hypotheses (per feature).

std::size_t opentl::core::cvdata::T_MEAS_FEAT::getNMaxMeas ( ) const [inline]

Get number of allocated features.

std::size_t opentl::core::cvdata::T_MEAS_FEAT::getNMeas ( ) const [inline]

Get number of available features.

void opentl::core::cvdata::T_MEAS_FEAT::resizeAll	(	T_MEAS_FEAT_TYPE	measType,
		std::size_t	nMaxMeas_,
		std::size_t	featElementDim_,
		std::size_t	jacSize_ = `0`,
		std::size_t	nMaxAssoc_ = `1`
	)			`[inline]`

Resizes all vectors within structure and sets internal variables accordingly.

Parameters:

	measType	Type of measurement: scalar or vector residual, with single or multiple association hypotheses.
	nMaxMeas_	Maximum number of features (in order to pre-allocate memory). The actual number of stored measurements is given by nMeas.
	featElementDim_	Length of one feature element (in case of vector data).
	jacSize_	Column-size of the Jacobian matrix (usually the state, or pose dimension). The row-size is featElementDim.
	nMaxAssoc_	Maximum number of association hypotheses (in case of multiple hypotheses, in order to pre-allocate the vectors).

void opentl::core::cvdata::T_MEAS_FEAT::setNMeas ( std::size_t nMeas_ ) [inline]

Set number of available features (it must be less or equal than the number of allocated features).

Member Data Documentation

double opentl::core::cvdata::T_MEAS_FEAT::alpha

Probability of missing detection (0 associated measurements) for each feature.

std::vector<std::vector<double> > opentl::core::cvdata::T_MEAS_FEAT::E_mHypo

E_mHypo = vector of measurement residuals (or innovations) associated to each pair (h,z), with possibly multiple hypotheses NOTE: The actual number of valid residuals is given by the respective entry in nAssociations (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: local feature index
inner (2nd) [0,nAssociations[f]]: association hypotheses (NOTE: the actual number may be less than nMaxAssoc).

std::vector<double> opentl::core::cvdata::T_MEAS_FEAT::E_sHypo

E_sHypo = vector of image residuals associated to each h, with either 0 or 1 association hypothesis. NOTE: The respective entry in nAssociations decides for the validity of this measurement (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: encodes the set of data in a single vector (in case of missing detection, the respective data are invalid).

std::vector<double> opentl::core::cvdata::T_MEAS_FEAT::h

h = vector of expected image features (obtained for example by re-projection from object space). Dimensions:

outer (1st) [0,nMeas*featElemDim]: local feature index

std::vector<double> opentl::core::cvdata::T_MEAS_FEAT::invCov

invCov = inverse of measurement covariances, for each feature. ATTENTION: invCov encodes a sequence of (nMeas) square matrices, each one of size (featElementDim * featElementDim), into a single vector. Therefore, the total dimension of invCov is (nMeas*featElementDim*featElementDim).

std::vector<int> opentl::core::cvdata::T_MEAS_FEAT::nAssociations

nAssociations = vector of integers, giving the actual number of associated hypotheses per feature, between [0,nMaxAssoc].

std::vector<std::vector<double> > opentl::core::cvdata::T_MEAS_FEAT::transpJac

transpJac = transposed Jacobians of h. See the jacSize parameter in resizeAll(). Dimensions:

outer (1st) [0,jacSize]: state variable index
inner (2nd) [0,nMeas*featElemDim]: local feature index (in case of missing detection, data in this vector are invalid).

Example: if the state dimension is 6, and you have 100 features, each one with a descriptor of dimension 3, J is (300 x 6) and transpJac will be (6 x 300).

std::vector<std::vector<double> > opentl::core::cvdata::T_MEAS_FEAT::z_mHypo

z_mHypo = vector of image features associated to each h, with possibly multiple hypotheses NOTE: The actual number of valid associations is given by the respective entry in nAssociations (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: local feature index
inner (2nd) [0,nAssociations[f]]: association hypotheses (NOTE: the actual number may be less than nMaxAssoc).

std::vector<double> opentl::core::cvdata::T_MEAS_FEAT::z_sHypo

z_sHypo = vector of image features associated to each h, with either 0 or 1 association hypothesis. NOTE: The respective entry in nAssociations decides for the validity of this measurement (0 = missing detection). Dimensions:

outer (1st) [0,nMeas*featElemDim]: encodes the set of data in a single vector (in case of missing detection, the respective data are invalid).


Public Types
Public Member Functions
std::size_t	getFeatElementDim () const
	Get size/length of a single feature element.
std::size_t	getJacSize () const
	Get the column-size of Jacobian matrices.
T_MEAS_FEAT_TYPE	getMeasType () const
	Return type of measurement.
std::size_t	getNMaxAssoc () const
	Get number of maximum association hypotheses (per feature).
std::size_t	getNMaxMeas () const
	Get number of allocated features.
std::size_t	getNMeas () const
	Get number of available features.
void	resizeAll (T_MEAS_FEAT_TYPE measType, std::size_t nMaxMeas_, std::size_t featElementDim_, std::size_t jacSize_=0, std::size_t nMaxAssoc_=1)
	Resizes all vectors within structure and sets internal variables accordingly.
void	setNMeas (std::size_t nMeas_)
	Set number of available features (it must be less or equal than the number of allocated features).
	T_MEAS_FEAT (std::size_t nMaxMeas_=1, std::size_t featElementDim_=0, std::size_t jacSize_=0, std::size_t nMaxAssoc_=1, double alpha_=0.0, T_MEAS_FEAT_TYPE measType=VECTOR_RES_SINGLE_HYPO)
	Constructor.
Public Attributes
double	alpha
	Probability of missing detection (0 associated measurements) for each feature.
std::vector< std::vector < double > >	E_mHypo
	E_mHypo = vector of measurement residuals (or innovations) associated to each pair (h,z), with possibly multiple hypotheses NOTE: The actual number of valid residuals is given by the respective entry in nAssociations (0 = missing detection). Dimensions: outer (1st) [0,nMeas*featElemDim]: local feature index inner (2nd) [0,nAssociations[f]]: association hypotheses (NOTE: the actual number may be less than nMaxAssoc).
std::vector< double >	E_sHypo
	E_sHypo = vector of image residuals associated to each h, with either 0 or 1 association hypothesis. NOTE: The respective entry in nAssociations decides for the validity of this measurement (0 = missing detection). Dimensions: outer (1st) [0,nMeas*featElemDim]: encodes the set of data in a single vector (in case of missing detection, the respective data are invalid).
std::vector< double >	h
	h = vector of expected image features (obtained for example by re-projection from object space). Dimensions: outer (1st) [0,nMeas*featElemDim]: local feature index
std::vector< double >	invCov
	invCov = inverse of measurement covariances, for each feature. ATTENTION: invCov encodes a sequence of (nMeas) square matrices, each one of size (featElementDim * featElementDim), into a single vector. Therefore, the total dimension of invCov is (nMeasfeatElementDimfeatElementDim).
std::vector< int >	nAssociations
	nAssociations = vector of integers, giving the actual number of associated hypotheses per feature, between [0,nMaxAssoc].
std::vector< std::vector < double > >	transpJac
	transpJac = transposed Jacobians of h. See the jacSize parameter in resizeAll(). Dimensions: outer (1st) [0,jacSize]: state variable index inner (2nd) [0,nMeas*featElemDim]: local feature index (in case of missing detection, data in this vector are invalid).
std::vector< std::vector < double > >	z_mHypo
	z_mHypo = vector of image features associated to each h, with possibly multiple hypotheses NOTE: The actual number of valid associations is given by the respective entry in nAssociations (0 = missing detection). Dimensions: outer (1st) [0,nMeas*featElemDim]: local feature index inner (2nd) [0,nAssociations[f]]: association hypotheses (NOTE: the actual number may be less than nMaxAssoc).
std::vector< double >	z_sHypo
	z_sHypo = vector of image features associated to each h, with either 0 or 1 association hypothesis. NOTE: The respective entry in nAssociations decides for the validity of this measurement (0 = missing detection). Dimensions: outer (1st) [0,nMeas*featElemDim]: encodes the set of data in a single vector (in case of missing detection, the respective data are invalid).