lti::linearRegressionTracking Class Reference

With this tracking algorithm it is possible to track the movement of an object together with scaling and rotation in the image plane. More...

#include <ltiLinearRegressionTracking.h>

Inheritance diagram for lti::linearRegressionTracking:

[legend]

Collaboration diagram for lti::linearRegressionTracking:

[legend]

List of all members.

Classes
class	parameters
	the parameters for the class linearRegressionTracking More...
Public Member Functions
	linearRegressionTracking ()
	linearRegressionTracking (const parameters &par)
	linearRegressionTracking (const linearRegressionTracking &other)
virtual	~linearRegressionTracking ()
virtual const char *	getTypeName () const
bool	learnDisplacement (channel &referenceImage, rectangle imageSection)
bool	learnDisplacementRGB (image &referenceImage, rectangle imageSection)
void	getDisplacement (channel &theImage, int xpos, int ypos, double &dx, double &dy) const
void	getDisplacementRGB (image &theImage, int xpos, int ypos, double &dx, double &dy) const
bool	learnTracking (channel &referenceImage, rectangle imageSection)
bool	learnTrackingRGB (image &referenceImage, rectangle imageSection)
void	getDispRotScale (channel &theImage, double xpos, double ypos, double alpha, double scale, double &dx, double &dy, double &dAlpha, double &dScale) const
void	getDispRotScaleRGB (image &theImage, double xpos, double ypos, double alpha, double scale, double &dx, double &dy, double &dAlpha, double &dScale) const
linearRegressionTracking &	copy (const linearRegressionTracking &other)
linearRegressionTracking &	operator= (const linearRegressionTracking &other)
virtual functor *	clone () const
const parameters &	getParameters () const
virtual bool	read (ioHandler &handler, const bool complete=true)
virtual bool	write (ioHandler &handler, const bool complete=true) const
void	getReferenceSection (vector< double > &referenceSection, int &w, int &h) const
void	setReferenceSection (vector< double > &referenceSection, int w, int h)
Protected Attributes
vector< double >	referenceValues
int	width

---

Detailed Description

With this tracking algorithm it is possible to track the movement of an object together with scaling and rotation in the image plane.

It uses the difference between the reference image of the object and the image values at the current position to estimate the correction of the geometrical parameters. This relation is calculated in a training phase using linear regression. To achieve stable tracking results, the image values in the background should be similar to those in the training image. It is also required that the movement of the object between subsequent frames is not too big, there should be still an overlap of the object areas at the two positions.

---

Constructor & Destructor Documentation

lti::linearRegressionTracking::linearRegressionTracking

(

)

default constructor

lti::linearRegressionTracking::linearRegressionTracking

(

const parameters &

par

)

Construct a functor using the given parameters.

lti::linearRegressionTracking::linearRegressionTracking

(

const linearRegressionTracking &

other

)

copy constructor

Parameters:

other

the object to be copied

virtual lti::linearRegressionTracking::~linearRegressionTracking

(

)

[virtual]

destructor

---

Member Function Documentation

virtual functor* lti::linearRegressionTracking::clone

(

)

const [virtual]

returns a pointer to a clone of this functor.

Reimplemented from lti::linearRegression< double >.

linearRegressionTracking& lti::linearRegressionTracking::copy

(

const linearRegressionTracking &

other

)

copy data of "other" functor.

Parameters:

other

the functor to be copied

Returns:

a reference to this functor object

Reimplemented from lti::functor.

void lti::linearRegressionTracking::getDisplacement	(	channel &	theImage,
		int	xpos,
		int	ypos,
		double &	dx,
		double &	dy
	)			const

estimates the object displacement from the difference of current and reference channel values

Parameters:

	theImage	current frame channel
	xpos	current x-position of left upper edge of object
	ypos	current y-position of left upper edge of object
	dx	estimated displacement in x-direction
	dy	estimated displacement in y-direction

void lti::linearRegressionTracking::getDisplacementRGB	(	image &	theImage,
		int	xpos,
		int	ypos,
		double &	dx,
		double &	dy
	)			const

estimates the object displacement from the difference of current and reference RGB values

Parameters:

	theImage	current frame channel
	xpos	current x-position of left upper edge of object
	ypos	current y-position of left upper edge of object
	dx	estimated displacement in x-direction
	dy	estimated displacement in y-direction

void lti::linearRegressionTracking::getDispRotScale	(	channel &	theImage,
		double	xpos,
		double	ypos,
		double	alpha,
		double	scale,
		double &	dx,
		double &	dy,
		double &	dAlpha,
		double &	dScale
	)			const

estimates the object displacement, scaling and rotation from the difference of current and reference channel values.

Use more than one iteration per frame for best results.

Parameters:

	theImage	current frame channel
	xpos	current x-position of object center
	ypos	current y-position of object center
	alpha	current object angle
	scale	current object scale
	dx	estimated displacement in x-direction
	dy	estimated displacement in y-direction
	dAlpha	estimated angle difference
	dScale	estimated scale difference

void lti::linearRegressionTracking::getDispRotScaleRGB	(	image &	theImage,
		double	xpos,
		double	ypos,
		double	alpha,
		double	scale,
		double &	dx,
		double &	dy,
		double &	dAlpha,
		double &	dScale
	)			const

estimates the object displacement, scaling and rotation from the difference of current and reference RGB values.

Use more than one iteration per frame for best results.

Parameters:

	theImage	current frame image
	xpos	current x-position of object center
	ypos	current y-position of object center
	alpha	current object angle
	scale	current object scale
	dx	estimated displacement in x-direction
	dy	estimated displacement in y-direction
	dAlpha	estimated angle difference
	dScale	estimated scale difference

const parameters& lti::linearRegressionTracking::getParameters

(

)

const

returns used parameters

Reimplemented from lti::linearRegression< double >.

void lti::linearRegressionTracking::getReferenceSection	(	vector< double > &	referenceSection,
		int &	w,
		int &	h
	)			const

gets the values of the reference image section used for calculating the error vector

Parameters:

	referenceSection	contains either the grey-scale or RGB-values in a vector
	w	width of image section
	h	height of image section

virtual const char* lti::linearRegressionTracking::getTypeName

(

)

const [virtual]

returns the name of this type ("linearRegressionTracking")

Reimplemented from lti::linearRegression< double >.

bool lti::linearRegressionTracking::learnDisplacement	(	channel &	referenceImage,
		rectangle	imageSection
	)

training of x- and y-movement in the image plane using channel values

Parameters:

	referenceImage	reference channel to be used for training
	imageSection	rectangle that surrounds the object in the given image

bool lti::linearRegressionTracking::learnDisplacementRGB	(	image &	referenceImage,
		rectangle	imageSection
	)

training of x- and y-movement in the image plane using RGB values

Parameters:

	referenceImage	reference image to be used for training
	imageSection	rectangle that surrounds the object in the given image

bool lti::linearRegressionTracking::learnTracking	(	channel &	referenceImage,
		rectangle	imageSection
	)

training of x- and y-movement as well as scaling and rotation in the image plane using channel values

Parameters:

	referenceImage	reference channel to be used for training
	imageSection	rectangle that surrounds the object in the given image

bool lti::linearRegressionTracking::learnTrackingRGB	(	image &	referenceImage,
		rectangle	imageSection
	)

training of x- and y-movement as well as scaling and rotation in the image plane using RGB values

Parameters:

	referenceImage	reference channel to be used for training
	imageSection	rectangle that surrounds the object in the given image

linearRegressionTracking& lti::linearRegressionTracking::operator=

(

const linearRegressionTracking &

other

)

alias for copy member

Parameters:

other

the functor to be copied

Returns:

a reference to this functor object

Reimplemented from lti::functor.

virtual bool lti::linearRegressionTracking::read	(	ioHandler &	handler,
		const bool	complete = true
	)			[virtual]

reads this functor from the given handler.

Reimplemented from lti::linearRegression< double >.

void lti::linearRegressionTracking::setReferenceSection	(	vector< double > &	referenceSection,
		int	w,
		int	h
	)

sets the values of the reference image section used for calculating the error vector

Parameters:

	referenceSection	contains either the grey-scale or RGB-values in a vector
	w	width of image section
	h	height of image section

virtual bool lti::linearRegressionTracking::write	(	ioHandler &	handler,
		const bool	complete = true
	)			const [virtual]

writes this functor to the given handler.

Reimplemented from lti::linearRegression< double >.

---

Member Data Documentation

vector<double> lti::linearRegressionTracking::referenceValues [protected]

reference image values (grey level or RGB-values subsequently) of the image region which shall be tracked

int lti::linearRegressionTracking::width [protected]

width and height of the image region

---

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

• ltiLinearRegressionTracking.h