lti::kalmanTracker Class Reference

A tracker implementation making use of the class kalmanFilter. More...

#include <ltiKalmanTracker.h>

Inheritance diagram for lti::kalmanTracker:

[legend]

Collaboration diagram for lti::kalmanTracker:

[legend]

List of all members.

Classes
class	parameters
	The parameters for the class kalmanTracker. More...
Public Member Functions
	kalmanTracker (const int &stateDimensions=4)
	kalmanTracker (const kalmanTracker &other)
virtual	~kalmanTracker ()
virtual const char *	getTypeName () const
int	getSystemStateDimensions () const
bool	apply (const float &xMeasurement, const float &yMeasurement, float &xPredictionApriori, float &yPredictionApriori)
bool	apply (float &xPredictionApriori, float &yPredictionApriori)
kalmanTracker &	copy (const kalmanTracker &other)
virtual functor *	clone () const
const parameters &	getParameters (void) const
bool	updateParameters ()
void	setMeasurementNoiseCovariance (const matrix< float > &r)
const vector< float > &	getEstimate () const
const vector< float > &	getEstimateAposteriori () const
const matrix< float > &	getErrorCovarianceApriori () const
const matrix< float > &	getErrorCovarianceAposteriori () const
void	reset (void)
bool	isInitialized () const
Protected Member Functions
bool	performTimeUpdate ()
bool	performMeasurementUpdate (const float &xMeasurement, const float &yMeasurement)
Protected Attributes
kalmanFilter	kalman
bool	initialized
lti::vector< float >	measurement
lti::vector< float >	prediction

Detailed Description

A tracker implementation making use of the class kalmanFilter.

It tracks a 2D point, using a system state vector that contains x and y coordinate as well as x and y velocity (denoted by vx and vy) and, optionally, x and y acceleration (ax and ay) of this point. These state vectors correspond to the assumption of constant velocity and constant acceleration, respectively. The state vector is structured like this: (x,vx,y,vy), or (x,vx,y,vy,ax,ay). The measured x and y coordinate are given to the apply method, which returns the estimated x and y coordinates at the next time step.

These matrices (contained in the parameters) must be specified:

matrix<float> measurementNoiseCovariance ( $R$ ; 2x2)
matrix<float> processNoiseCovariance ( $Q$ ; 4x4 / 6x6)

These matrices (contained in the parameters) should be specified, but are initialized to default values that should be sufficient for a first try:

vector<float> initialSystemState ( $\hat{x}^{-}$ ; 4 / 6)
matrix<float> initialErrorCovariance ( $P^{-}$ and $P$ ; 4x4 / 6x6)

The parantheses contain notation and dimensionality of the matrices for the constant velocity / constant acceleration case.

Constructor & Destructor Documentation

lti::kalmanTracker::kalmanTracker ( const int & stateDimensions = 4 )

constructor for n-dimensional system state (n may be 4 or 6).

default is 4.

lti::kalmanTracker::kalmanTracker ( const kalmanTracker & other )

copy constructor

Parameters:

other

the object to be copied

virtual lti::kalmanTracker::~kalmanTracker ( ) [virtual]

destructor

Member Function Documentation

bool lti::kalmanTracker::apply	(	float &	xPredictionApriori,
		float &	yPredictionApriori
	)

Perform only a time update (i.e.

no new measurement available), resulting in a prediction of x and y coordinates for the next time step, which are then returned.

bool lti::kalmanTracker::apply	(	const float &	xMeasurement,
		const float &	yMeasurement,
		float &	xPredictionApriori,
		float &	yPredictionApriori
	)

Perform a measurement update and then a time update, resulting in a prediction of x and y coordinates for the next time step, which are then returned.

virtual functor* lti::kalmanTracker::clone ( ) const [virtual]

returns a pointer to a clone of this functor.

Implements lti::functor.

kalmanTracker& lti::kalmanTracker::copy ( const kalmanTracker & other )

copy data of "other" functor.

Parameters:

other

the functor to be copied

Returns:: a reference to this functor object

Reimplemented from lti::functor.

const matrix<float>& lti::kalmanTracker::getErrorCovarianceAposteriori ( ) const

Get the a posteriori error covariance ( $P$ ).

const matrix<float>& lti::kalmanTracker::getErrorCovarianceApriori ( ) const

Get the a priori error covariance ( $P^{-}$ ).

const vector<float>& lti::kalmanTracker::getEstimate ( ) const

Get the (a priori) system state estimate ( $\hat{x}^{-}$ ).

This is the prediction one would normally be interested in.

const vector<float>& lti::kalmanTracker::getEstimateAposteriori ( ) const

Get the a posteriori system state estimate ( $\hat{x}$ ).

const parameters& lti::kalmanTracker::getParameters ( void ) const

returns used parameters

Reimplemented from lti::functor.

int lti::kalmanTracker::getSystemStateDimensions ( ) const

Return the number of system state dimensions (currently either 4 or 6).

virtual const char* lti::kalmanTracker::getTypeName ( ) const [virtual]

returns the name of this type ("kalmanTracker")

Reimplemented from lti::functor.

bool lti::kalmanTracker::isInitialized ( ) const

Tell if this tracker is initialized, i.e.

if any of the apply methods has been called after the creation of this object or a call to reset().

bool lti::kalmanTracker::performMeasurementUpdate	(	const float &	xMeasurement,
		const float &	yMeasurement
	)			`[protected]`

Perform measurement update.

bool lti::kalmanTracker::performTimeUpdate ( ) [protected]

Perform time update.

void lti::kalmanTracker::reset ( void )

Reset the filter state ( $x$ ) and error covariance ( $P$ ).

The tracker is left in an un-initialized state.

void lti::kalmanTracker::setMeasurementNoiseCovariance ( const matrix< float > & r )

set a new measurement noise covariance, e.g.

because future measurements are less/more reliable (overlapping starts/ends)

bool lti::kalmanTracker::updateParameters ( ) [virtual]

set parameters of this Kalman tracker

Reimplemented from lti::functor.

Member Data Documentation

bool lti::kalmanTracker::initialized [protected]

Initialization status of this tracker.

Set to false after object creation or call to reset(). Set to true after the first update, i.e. call to an apply method.

kalmanFilter lti::kalmanTracker::kalman [protected]

The kalmanFilter used by this tracker.

lti::vector<float> lti::kalmanTracker::measurement [protected]

This temporary variable wraps two floats (the x/y measurement) in a vector.

lti::vector<float> lti::kalmanTracker::prediction [protected]

This temporary variable wraps two floats (the x/y prediction) in a vector.

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

ltiKalmanTracker.h

lti::kalmanTracker Class Reference

Classes

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation