This class implements a tracker determining a similarity transformation between two successive video frames. More...

Public Types
enum	TrackerConfidence { TC_NONE , TC_BAD , TC_MODERATE , TC_GOOD , TC_VERY_GOOD }
	Definition of individual confidence values. More...

enum	RegionTextureness { RT_UNKNOWN , RT_LOW , RT_MODERATE , RT_HIGH }
	Definition of individual textureness qualities. More...

Public Member Functions
	SimilarityTracker ()=default
	Creates a new tracker object. More...

bool	determineSimilarity (const Frame &yFrame, const CV::PixelBoundingBox &previousSubRegion, SquareMatrix3 currentTprevious=nullptr, Vector2 translation=nullptr, Scalar rotation=nullptr, Scalar scale=nullptr, const Vector2 &predictedTranslation=Vector2(0, 0), TrackerConfidence trackerConfidence=nullptr, RegionTextureness regionTextureness=nullptr, Worker *worker=nullptr)
	Determines the similarity between two successive video frame within a specified sub-region. More...

const CV::FramePyramid &	currentFramePyramid () const
	Returns the frame pyramid of the most recent frame. More...

const CV::FramePyramid &	keyFramePyramid () const
	Returns the frame pyramid of one of the previous frames (the current key-frame). More...

void	reset ()
	Resets the similarity tracker. More...

Static Public Member Functions
static float	combinedConfidence (const TrackerConfidence trackerConfidence, const RegionTextureness regionTextureness)
	Calculate an overall confidence values based on a tracker confidence and a region textureness. More...

Static Protected Member Functions
static bool	determineFeaturePoints (const CV::FramePyramid &yFramePyramid, const CV::PixelBoundingBox &subRegion, const size_t minimalFeaturePoints, const size_t desiredFeaturePoints, const size_t maximalFeaturePoints, Vectors2 &featurePoints, unsigned int &usedLayerIndex, RegionTextureness &regionTextureness, Worker *worker=nullptr)
	Detects feature points in given frame for which a frame pyramid exists. More...

static bool	determineSimilarityTransformation (const CV::FramePyramid &yPreviousFramePyramid, const CV::FramePyramid &yCurrentFramePyramid, const Vectors2 &previousPoints, const Vectors2 &roughCurrentPoints, RandomGenerator &randomGenerator, SquareMatrix3 &currentTprevious, Indices32 &validCorrespondences, const unsigned int coarsestLayerRadius, const unsigned int subPixelIterations, Worker *worker=nullptr)
	Determines the similarity transformation between two successive frames. More...

Protected Attributes
CV::FramePyramid	keyFramePyramid_
	The frame pyramid of the key frame, one of the previous frames. More...

CV::FramePyramid	currentFramePyramid_
	The frame pyramid of the current frame. More...

RandomGenerator	randomGenerator_
	The random generator object. More...

SquareMatrix3	previous_T_key_ = SquareMatrix3(true)
	The similarity since 'previousFramePyramid_' have been updated the last time, defined in the resolution of the usage pyramid layer. More...

Vectors2	keyFramePoints_
	The image points located in 'previousFramePyramid_', defined in the usage pyramid layer. More...

unsigned int	keyFramePointsLayerIndex_ = (unsigned int)(-1)
	The frame pyramid layer index in which 'keyFramePoints_' have been determined, -1 if invalid. More...

Detailed Description

This class implements a tracker determining a similarity transformation between two successive video frames.

A similarity transformation has four degrees of freedom and contains a rotation, a scale, and a 2D translation (in x- and y- direction) within the image domain.
The 3x3 matrix representing the similarity transformation has the following layout:

| a  -b  tx |
| b   a  ty |
| 0   0   1 |

Member Enumeration Documentation

◆ RegionTextureness

enum Ocean::Tracking::Point::SimilarityTracker::RegionTextureness

Definition of individual textureness qualities.

Enumerator
RT_UNKNOWN	The textureness is unknown.
RT_LOW	The textureness is low (an almost homogeneous region).
RT_MODERATE	The textureness is moderate.
RT_HIGH	The textureness is high (an almost heterogeneous region).

◆ TrackerConfidence

enum Ocean::Tracking::Point::SimilarityTracker::TrackerConfidence

Definition of individual confidence values.

Enumerator
TC_NONE	No similarity could be determined.
TC_BAD	The determined similarity may have a bad quality.
TC_MODERATE	The determined similarity has a moderate quality.
TC_GOOD	The determined similarity has a good quality.
TC_VERY_GOOD	The determined similarity has a very good quality and can be trusted regardless.

Constructor & Destructor Documentation

◆ SimilarityTracker()

Ocean::Tracking::Point::SimilarityTracker::SimilarityTracker ( )

default

Creates a new tracker object.

Member Function Documentation

◆ combinedConfidence()

float Ocean::Tracking::Point::SimilarityTracker::combinedConfidence	(	const TrackerConfidence	trackerConfidence,
		const RegionTextureness	regionTextureness
	)

inlinestatic

Calculate an overall confidence values based on a tracker confidence and a region textureness.

Parameters

trackerConfidence	The confidence value of the tracker
regionTextureness	The textureness of the sub-region in which the tracker was applied

Returns: The resulting overall confidence value, with range [0, 1]

◆ currentFramePyramid()

const CV::FramePyramid & Ocean::Tracking::Point::SimilarityTracker::currentFramePyramid ( ) const

inline

Returns the frame pyramid of the most recent frame.

Returns: Frame pyramid of current frame, if any

◆ determineFeaturePoints()

static bool Ocean::Tracking::Point::SimilarityTracker::determineFeaturePoints	(	const CV::FramePyramid &	yFramePyramid,
		const CV::PixelBoundingBox &	subRegion,
		const size_t	minimalFeaturePoints,
		const size_t	desiredFeaturePoints,
		const size_t	maximalFeaturePoints,
		Vectors2 &	featurePoints,
		unsigned int &	usedLayerIndex,
		RegionTextureness &	regionTextureness,
		Worker *	worker = `nullptr`
	)

staticprotected

Detects feature points in given frame for which a frame pyramid exists.

The points may be determined in a lower image resolution if enough features can be found.

Parameters

yFramePyramid	The frame pyramid of the frame in which the feature points will be determined, with pixel format FORMAT_Y8, must be valid with at least one layer
subRegion	The sub-region inside the frame in which feature points will be determined, specified in the domain of the finest image resolution, must be valid
minimalFeaturePoints	The minimal number of feature points that must be detected so that the function succeeds, with range [1, infinity)
desiredFeaturePoints	The desired number of feature points that should be detected in an ideal case, with range [minimalFeaturePoints, infinity)
maximalFeaturePoints	The maximal number of feature points that will be detected, with range [desiredFeaturePoints, infinity)
featurePoints	The resulting detected feature points, defined in the domain of the pyramid layer as returned by 'usedLayerIndex'
usedLayerIndex	The pyramid layer in which the resulting feature points have been detected, with range [0, framePyramid.layers() - 1]
regionTextureness	The resulting textureness of the sub-region (which is based on the numbers of found feature etc.)
worker	Optional worker object to distribute the computation

Returns: True, if at least 'minimalFeaturePoints' could be determined

◆ determineSimilarity()

bool Ocean::Tracking::Point::SimilarityTracker::determineSimilarity	(	const Frame &	yFrame,
		const CV::PixelBoundingBox &	previousSubRegion,
		SquareMatrix3 *	currentTprevious = `nullptr`,
		Vector2 *	translation = `nullptr`,
		Scalar *	rotation = `nullptr`,
		Scalar *	scale = `nullptr`,
		const Vector2 &	predictedTranslation = `Vector2(0, 0)`,
		TrackerConfidence *	trackerConfidence = `nullptr`,
		RegionTextureness *	regionTextureness = `nullptr`,
		Worker *	worker = `nullptr`
	)

Determines the similarity between two successive video frame within a specified sub-region.

Parameters

yFrame	The 8bit grayscale frame for which the similarity (in relation to the previous frame) will be determined, with image dimension [40, infinity)x[40, infinity) must be valid
previousSubRegion	The sub-region (in the previous frame) in which feature points will be determined and tracked to from the previous frame to the current frame, must be valid, must not lie outside the frame
currentTprevious	Optional resulting 3x3 matrix containing the entire similarity transformation: currentPoint = similarity * previousPoint
translation	Optional resulting translation between the current and the previous video frame, in pixel, with range (-infinity, infinity)x(-infinity, infinity)
rotation	Optional resulting rotation between the current and the previous video frame, in radian, with range [-PI, PI]
scale	Optional resulting scale between the current and the previous video frame, with range (0, infinity)
predictedTranslation	The predicted translation between the previous frame and the current frame if known, may be based on e.g., a gyro motion
trackerConfidence	Optional resulting confidence value of the tracker
regionTextureness	Optional resulting textureness of the provided sub-region
worker	Optional worker object to distribute the computation

Returns: True, if succeeded, use trackerConfidence to check whether the resulting tracking data is valid

◆ determineSimilarityTransformation()

static bool Ocean::Tracking::Point::SimilarityTracker::determineSimilarityTransformation	(	const CV::FramePyramid &	yPreviousFramePyramid,
		const CV::FramePyramid &	yCurrentFramePyramid,
		const Vectors2 &	previousPoints,
		const Vectors2 &	roughCurrentPoints,
		RandomGenerator &	randomGenerator,
		SquareMatrix3 &	currentTprevious,
		Indices32 &	validCorrespondences,
		const unsigned int	coarsestLayerRadius,
		const unsigned int	subPixelIterations,
		Worker *	worker = `nullptr`
	)

staticprotected

Determines the similarity transformation between two successive frames.

Parameters

yPreviousFramePyramid	The frame pyramid of the previous frame in which the given 'previousPoints' are located, with pixel format FORMAT_Y8, must be valid with at least one layer
yCurrentFramePyramid	The frame pyramid of the current frame to which the previous points will be tracked, must have the same format and layout as 'yPreviousFramePyramid'
previousPoints	The previous feature points defined within the domain of the finest pyramid layer of 'yPreviousFramePyramid', at least one
roughCurrentPoints	Optional the already known rough locations of the previous points with the current frame, one for each previous point, otherwise an empty vector
randomGenerator	The object to be used to generate random numbers
currentTprevious	The resulting similarity transformation matching with the resolution of f the finest pyramid layer, transforming points defined in the previous frame to points in the current frame: (currentPoint = currentTprevious * previousPoint)
validCorrespondences	The resulting indices of all previous feature points that have been used for determination of the similarity transformation
coarsestLayerRadius	The search radius on the coarsest layer in pixel, with range [2, infinity)
subPixelIterations	Number of sub-pixel iterations that will be applied, each iteration doubles the sub-pixel accuracy, with range [1, infinity)
worker	Optional worker object to distribute the computation

Returns: True, if succeeded

◆ keyFramePyramid()

const CV::FramePyramid & Ocean::Tracking::Point::SimilarityTracker::keyFramePyramid ( ) const

inline

Returns the frame pyramid of one of the previous frames (the current key-frame).

Returns: Frame pyramid of previous frame, if any

◆ reset()

void Ocean::Tracking::Point::SimilarityTracker::reset ( )

inline

Resets the similarity tracker.

Field Documentation

◆ currentFramePyramid_

CV::FramePyramid Ocean::Tracking::Point::SimilarityTracker::currentFramePyramid_

protected

The frame pyramid of the current frame.

◆ keyFramePoints_

Vectors2 Ocean::Tracking::Point::SimilarityTracker::keyFramePoints_

protected

The image points located in 'previousFramePyramid_', defined in the usage pyramid layer.

◆ keyFramePointsLayerIndex_

unsigned int Ocean::Tracking::Point::SimilarityTracker::keyFramePointsLayerIndex_ = (unsigned int)(-1)

protected

The frame pyramid layer index in which 'keyFramePoints_' have been determined, -1 if invalid.

◆ keyFramePyramid_

CV::FramePyramid Ocean::Tracking::Point::SimilarityTracker::keyFramePyramid_

protected

The frame pyramid of the key frame, one of the previous frames.

◆ previous_T_key_

SquareMatrix3 Ocean::Tracking::Point::SimilarityTracker::previous_T_key_ = SquareMatrix3(true)

protected

The similarity since 'previousFramePyramid_' have been updated the last time, defined in the resolution of the usage pyramid layer.

◆ randomGenerator_

RandomGenerator Ocean::Tracking::Point::SimilarityTracker::randomGenerator_

protected

The random generator object.

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

SimilarityTracker.h

Public Types

Public Member Functions

Static Public Member Functions

Static Protected Member Functions

Protected Attributes

Detailed Description

Member Enumeration Documentation

◆ RegionTextureness

◆ TrackerConfidence

Constructor & Destructor Documentation

◆ SimilarityTracker()

Member Function Documentation

◆ combinedConfidence()

◆ currentFramePyramid()

◆ determineFeaturePoints()

◆ determineSimilarity()

◆ determineSimilarityTransformation()

◆ keyFramePyramid()

◆ reset()

Field Documentation

◆ currentFramePyramid_

◆ keyFramePoints_

◆ keyFramePointsLayerIndex_

◆ keyFramePyramid_

◆ previous_T_key_

◆ randomGenerator_