Analytical line track fitting processor for LUIZA. More...

#include <FindLineTrajectory.h>

Inheritance diagram for luiza::FindLineTrajectory:

Public Member Functions
virtual Processor *	newProcessor ()
	Returns a new instance of FindLineTrajectory. More...

	FindLineTrajectory ()
	Default constructor.

virtual void	init ()
	Called at the job beginning. More...

virtual void	startAnalysis (GloriaDataContainer *)
	Called before first data processing loop. More...

virtual void	processData (GloriaDataContainer *data)
	Called for every data processing loop. More...

virtual void	check (GloriaDataContainer *data)
	Called for every loop - right after processData() call. More...

virtual void	concludeAnalysis (GloriaDataContainer *)
	Called after all data processing loops. More...

virtual void	end ()
	Called after data processing for clean up. More...

Public Member Functions inherited from luiza::Processor
	Processor (const std::string &typeName)
	Default constructor. More...

virtual	~Processor ()
	Destructor.

virtual const std::string &	type () const
	Return type name for the processor (as set in constructor).

virtual const std::string &	name () const
	Return name of this processor.

virtual const std::string &	logLevelName () const
	Return name of the local verbosity level of this processor - "" if not set.

virtual StringParameters *	parameters ()
	Return parameters defined for this Processor.

virtual void	printDescription ()
	Print information about this processor in ASCII steering file format.

virtual void	printDescriptionXML (std::ostream &stream=std::cout)
	Print information about this processor in XML steering file format.

template<class T >
void	printParameters ()
	Print the parameters and their values depending on the given verbosity level.

void	printParameters ()
	Print the parameters and their values with verbosity level MESSAGE.

const std::string &	description ()
	Description of processor.

bool	isFirstEvent ()
	True if first event in processEvent(evt) - use this e.g. to initialize histograms etc.

GloriaCollectionType	getGloriaInType (const std::string &colName)
	Return the input type for the collection colName. More...

GloriaCollectionType	getGloriaOutType (const std::string &colName)
	Return the output type for the collection colName. More...

bool	isInputCollectionName (const std::string &parameterName)
	Check if parameter defines collection. More...

bool	isOutputCollectionName (const std::string &parameterName)
	True if the given parameter defines an LCIO output collection */.

Protected Member Functions
bool	initFit (GloriaDataContainer *)
	Prepare fit environment and calculate nominal fit parameters.

void	endFit ()
	Clean fit environment.

double	MatrixFit ()
	Find track. More...

double	SingleFit ()
	Find track assuming nominal errors. More...

double	NominalFit ()
	Find track on all frames. More...

int	DoAnalFit (double pos, double err, double velocity=0.)

double	GetFitChi2 ()
	Calculate $\chi^{2}$ of the fit. More...

int	GaussjSolve (double alfa, double beta, int n)
	Solve matrix equation.

Protected Member Functions inherited from luiza::Processor
void	setReturnValue (bool val)
	Set the return value for this processor - typically at end of processEvent(). More...

void	setReturnValue (const std::string &name, bool val)
	Set a named return value for this processor - typically at end of processEvent() More...

template<class T >
void	registerProcessorParameter (const std::string &name, const std::string &description, T &parameter, const T &defaultVal, int setSize=0)
	Register a steering variable for this processor - call in constructor of processor. More...

void	registerInputCollection (const GloriaCollectionType &type, const std::string &name, const std::string &description, std::string &parameter, const std::string &defaultVal, int setSize=0)
	Register parameter describing input collection. More...

void	registerOutputCollection (const GloriaCollectionType &type, const std::string &name, const std::string &description, std::string &parameter, const std::string &defaultVal, int setSize=0)
	Register parameter describing output collection. More...

void	registerInputCollections (const GloriaCollectionType &type, const std::string &name, const std::string &description, StringVec &parameter, const StringVec &defaultVal, int setSize=0)
	Register parameter describing input collections. More...

template<class T >
void	registerOptionalParameter (const std::string &name, const std::string &description, T &parameter, const T &defaultVal, int setSize=0)
	Same as registerProcessorParameter except that the parameter is optional. More...

bool	parameterSet (const std::string &name)
	Tests whether the parameter has been set in the steering file.

template<class T >
void	message (const std::string &message) const
	Print message according to verbosity level. More...

template<class T >
void	message (const std::basic_ostream< char, std::char_traits< char > > &m) const
	Similar to message(const std::string& message) More...

std::stringstream &	log () const
	Returns an empty stringstream that is used by the message method.

virtual void	setProcessorParameters (StringParameters *parameters)
	Allow friend class CCProcessor to change/reset processor parameters.

virtual void	updateParameters ()
	Allow friend class CCProcessor to update processor parameters.

virtual void	updateTelescopeParameters (std::string name)
	Allow to update processor parameters for given telescope name. More...

virtual void	updateTelescopeParameters (GloriaFitsHeader *header)
	Update processor parameters based on FITS header. More...

Protected Attributes
std::string	_inputCollectionName
	Name of the input object list collection.

std::string	_outputTrackCollectionName
	Name of the output collection for storing track parameters.

std::string	_outputPositionCollectionName
	Name of the collection for storing fitted object positions on subsequent frames.

int	_maxFrameObjects
	Maximum number of objects considered per.

double	_chi2Max
	Maximum $\chi^{2}$ for accepted track fit.

std::vector< std::string >	_positionColumns
	Names of the columns, where object positions are stored.

std::vector< float >	_positionErrors
	Assumed errors on object positions in Ra and Dec.

std::string	_timeRecord
	Name of header record storing frame time.

bool	_searchMultipleTracks
	Flag for searching multiple tracks.

bool	_allowAmbiguousObjects
	Allow same object to be used in more than track.

int	_maximumAmbiguousObjects
	Number of objects than can be shared by two tracks.

int	_allowMissingObservations
	Allowed number of frames with missing object on the track.

double	_missingObservationPenalty
	"Penalty" added to track $\chi^{2}$ for each frame without track object

bool	_useVelocityConstraint
	Flag for using expected direction and angular velocity constraint in the fit.

double	_velocitySpread
	Assumed spread of the fitted object velocity.

double	_raVelocity
	Assumed object sky velocity in Ra [deg/s].

double	_decVelocity
	Assumed object sky velocity in Dec [deg/s].

bool	_usePreselection
	Use additional cuts, based on measured object position on the first frame.

int	_nLoops
	Total number of processing loops.

int	_nFrames
	Total number of processed frames.

int	_nTracks
	Total number of reconstructed tracks.

Protected Attributes inherited from luiza::Processor
std::string	_description
	Describes what the processor does. Set in constructor.

Detailed Description

Analytical line track fitting processor for LUIZA.

This processor was adopted for fitting tracks to object positions on the sky reconstructed on subsequent frames. Input collection should contain object lists with objectc suspected to be satellites (based on shape, comparison with previous frame or matching with catalogue). There are usually multiple such objects in each frame, so main task of the processor is to look for the best track candidate considering all fit possibilities. All considered frames should be loaded in one loop.

Fit method: Track fitting is performed separately in RA vs time and DEC vs time. Object trajectory (position on each frame) is found by solving matrix equation resulting from $\chi^{2}$ minimum condition.

Track finding algorithm

Read object lists for frame set from input table collection and copy to local tables
Calculate number of fit hypothesis (including missing object possibility)
Search the list of fit hypotheses to find the one with best $\chi^{2}$ (including ``penalties'' for missing observation)
Accept the fit if $\chi^{2}$ is below threshold
Write fitted track to output collection; reconstructed object positions can also be stored to object collection.

Output: Fitted object position on the first frame as well as RA/Dec velocities are stored in the output track collection. Also fitted object positions at all frames (at corresponding times) can be stored in new collection or added to the input object list collection. Following parameters define input and output from the processor.

Parameters

InputCollectionName	Name of the input object list collection
OutputTrackCollectionName	Name of the output collection for storing track parameters (main output of the processor).
OutputPositionCollectionName	Name of the collection for storing fitted object positions on subsequent frames

Fit performance parameters

Parameters

MaxFrameObjects	Maximum number of objects considered per frame. Should be large (default is 100) for high track finding efficiency even for frames with high object multiplicities (but have a look at preformance issues below). For low values algorithm runs much faster (reduced number of hypothesis), but the efficiency is lower.
Chi2Max	Maximum $\chi^{2}$ for accepted track fit.
PositionColumns	Names of the columns, where object positions, to be used in the fit, are stored
PositionErrors	Assumed errors on object positions used in the fit (one number, or separate errors for Ra and Dec).
TimeRecord	Keyword of the header record storing frame time (assumed to be in seconds; default is TIME_UT)
SearchMultipleTracks	Flag for searching multiple tracks in lists with multiple objects. If false, only best (lowest $\chi^{2}$ ) track is taken.
AllowAmbiguousObjects	Allow same object to be used in more than one track.
MaximumAmbiguousObjects	Number of objects than can be shared by two tracks (when AllowAmbiguousObjects option is set)
UseVelocityConstraint	Flag for using expected direction and angular velocity constraint in the fit. Can improve the fit, also improves track searching for multiple objects.
RaVelocity	Assumed object sky velocity in Ra [deg/s]
DecVelocity	Assumed object sky velocity in Dec [deg/s]
VelocitySpread	Assumed spread of the fitted object velocity
AllowMissingObservations	Allowed number of frames with missing object on the track
MissingObservationPenalty	"Penalty" added to track $\chi^{2}$ for each frame without reconstructed object on the track
UsePreselection	Use additional cuts, based on measured object position on the first frame and the expected object track direction (velocities given when UseVelocityConstraint option is used) to constraint number of considered hit combinations and speed up the processing.

Performance issues: As described above, if multiple objects are found in sky images, the algorithm checks all object selection possibilities (all track hypothesis). This task is optimized to a large extent, but still track finding can be slow for large object multiplicities. Here are some suggestions on how to improve performance.

Use velocity constraint (set UseVelocityConstraint to true ), even if velocity spread is large. With velocity constraint first two frames are sufficent to recognize bad track hypothesis. Without velocity constraint at least 3 frames are needed.

Use track preselection based on velocity (set UsePreselection to true ). This helps a lot especially when the velocity is known (ie. for geostationary satellites).

Do not allow for missing observations (set AllowMissingObservations to 0). This reduces number of fit hypothesis and improves fit performance. However, track finding efficiency decreases.

Try to limit number of objects per plane, by using tighter object selection before calling track finder.

Author: A.F.Zarnecki, University of Warsaw, zarne.nosp@m.cki@.nosp@m.fuw.e.nosp@m.du.p.nosp@m.l

Member Function Documentation

void luiza::FindLineTrajectory::check ( GloriaDataContainer * data )

virtual

Called for every loop - right after processData() call.

Can be used to check processing and/or produce check plots.

Reimplemented from luiza::Processor.

void luiza::FindLineTrajectory::concludeAnalysis ( GloriaDataContainer * data )

virtual

Called after all data processing loops.

This method is called to finalize the analysis, after all input data was read (or required number of loops was reached). Analysis of data collected in permanent data collections is possible here.

Reimplemented from luiza::Processor.

int luiza::FindLineTrajectory::DoAnalFit	(	double *	pos,
		double *	err,
		double	velocity = `0.`
	)

protected

Fit track in one coordinate (Ra or Dec), taking into account object velocity

References _useVelocityConstraint, _velocitySpread, and GaussjSolve().

Referenced by initFit(), MatrixFit(), and SingleFit().

void luiza::FindLineTrajectory::end ( )

virtual

Called after data processing for clean up.

Used to release memory allocated in init() step

Reimplemented from luiza::Processor.

References _nFrames, _nLoops, and _nTracks.

double luiza::FindLineTrajectory::GetFitChi2 ( )

protected

Calculate $\chi^{2}$ of the fit.

Calculate $\chi^{2}$ of the fit taking into account measured object positions in Ra and Dec

References _decVelocity, _raVelocity, _useVelocityConstraint, and _velocitySpread.

Referenced by MatrixFit(), NominalFit(), and SingleFit().

void luiza::FindLineTrajectory::init ( )

virtual

Called at the job beginning.

This is executed only once in the whole execution.

Reimplemented from luiza::Processor.

References _inputCollectionName, _nFrames, _nLoops, _nTracks, _outputPositionCollectionName, _outputTrackCollectionName, _positionColumns, _positionErrors, luiza::Processor::name(), and luiza::Processor::parameters().

double luiza::FindLineTrajectory::MatrixFit ( )

protected

Find track.

Fit track in two coordinates (Ra and Dec) by solving two matrix equations and calculate $\chi^{2}$

References _decVelocity, _raVelocity, DoAnalFit(), and GetFitChi2().

Referenced by processData().

virtual Processor* luiza::FindLineTrajectory::newProcessor ( )

inlinevirtual

Returns a new instance of FindLineTrajectory.

This method returns an new instance of the this processor. It is called by Luiza execution framework and it shouldn't be called/used by the final user.

Returns: a new FindLineTrajectory

Implements luiza::Processor.

double luiza::FindLineTrajectory::NominalFit ( )

protected

Find track on all frames.

Fit track in two coordinates: Ra and Dec. When same position error is assumed and objects are found on all frames, same inverse matrix can be used for all events.

References _decVelocity, _raVelocity, _useVelocityConstraint, _velocitySpread, and GetFitChi2().

Referenced by processData().

void luiza::FindLineTrajectory::processData ( GloriaDataContainer * data )

virtual

Called for every data processing loop.

This is where most of the work (analysis) is done. Data is exchanged between processors via GloriaDataContainer

Reimplemented from luiza::Processor.

double luiza::FindLineTrajectory::SingleFit ( )

protected

Find track assuming nominal errors.

Fit track in two coordinates (Ra and Dec). When single position error is used, only one matrix equation has to be solved and the inverse matrix can be applied to the second equation.

References _decVelocity, _raVelocity, _useVelocityConstraint, _velocitySpread, DoAnalFit(), and GetFitChi2().

Referenced by processData().

void luiza::FindLineTrajectory::startAnalysis ( GloriaDataContainer * data )

virtual

Called before first data processing loop.

Can be used to initialize data collections required for analysis, eg. define permanent collections, or fill them with external data (from file, database or net). It is possible as GloriaDataContainer is already defined, contrary to init().

Reimplemented from luiza::Processor.

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

FindLineTrajectory.h
FindLineTrajectory.cc

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Member Function Documentation