Processor to find linear tracks using linear regression, based on a modified version of the track following algorithm. More...

#include <TrackFinderRectangularProcessor.h>

Inheritance diagram for marlintpc::TrackFinderRectangularProcessor:

Public Member Functions
virtual Processor *	newProcessor ()

virtual void	init ()
	Called at the begin of the job before anything is read. More...

virtual void	processRunHeader (lcio::LCRunHeader *run)
	Called for every run.

virtual void	processEvent (lcio::LCEvent *evt)
	Called for every event - the working horse.

virtual void	check (lcio::LCEvent *evt)

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

Protected Member Functions
void	zeroFlags (bool *IsIn)
	Set flags if hit is in a track to zero.

void	addTrackCandidateFlags ()
	Add the hits of the track candidate to the isInTrack flag array.

void	createNewIsInTrackFlags ()

IMPL::TrackerHitImpl	extrapolateNextHit (double y)
	Extrapolate the next expected hit on the track.

void	doRegression (EVENT::Track *trackCandidate)
	Calculate the track parameters from the hits. More...

Protected Attributes
std::string	_inputColName

std::string	_outputColName

int	_outputIsTransient
	give the status which is set to the transient flag of the output collection it is an int instead of a boolean, because the processor parameter cannot be a boolean

gear::PadRowLayout2D const *	_padLayout
	The pad layout as defined in GEAR.

int	_rowHeight

int	_minTrackHits
	Minimum number of hits on track.

int	_maxSkipRows
	Maximum number of missing hits.

float	_deltaX
	Maximum distances between predicted and measured hit.

float	_deltaY

float	_deltaZ

int	_nRun
	!!!!! This information should come from the conditions data! Implemented as processor parameter for first tests! More...

int	_nEvt
	The total number of events.

int	_verbosityLevel
	Flag about verbosity of the processor. More...

bool *	isInTrack
	Array of flags whether a hit is in a track.

bool *	isInTrackCandidate
	Array of flags whether a hit is in the current track candidate.

unsigned int	nHits
	Number of hits in the current event.

struct {
float a

float b

float c

float d

}	straightLine
	Straight line for linear regression. More...

Detailed Description

Processor to find linear tracks using linear regression, based on a modified version of the track following algorithm.

It assumes the hits being equidistantly spaced in y, which is the case for a TPC with rectangular pad layout (straight rows).

Input - Prerequisites

Collection of TrackerHits, sorted in y, x and z (in this order). A sorting function might be implemented in a later version or provided as a separate processor. For the time being the HitFinder has provide sorted hits.

Output

Collection of Tracks

Parameters

InputCollectionName	Name of input data collection (default: TPCHits)
OutputCollectionName	Name of output data collection (default: TPCSeedTracks)
SetOutputTransient	If not 0 the output collection is set transient (default: 0)

This processor is intended to work as a finder, the fitting is done by a more advanced fitting algorithm. In this case the output can be set transient, so the seed tracks are not written to disk.
As the finding algorithm performs a linear regression, which is equivalent to a $\chi^2$ minimisation for straight tracks, all track parameters are available and the track fitting can be skipped. In this case the output collection name should be changed to TPCTracks.

Parameters

MinTrackHits	Minimum number of hits on the track
MaxSkipRows	Maximum number of subsequently missing hits
DeltaX	Maximum x distances between predicted and measured hit
DeltaY	Maximum y distances between predicted and measured hit
DeltaZ	Maximum z distances between predicted and measured hit

Member Function Documentation

void marlintpc::TrackFinderRectangularProcessor::doRegression ( EVENT::Track * trackCandidate )

protected

Calculate the track parameters from the hits.

This is done by linear regression, which is equivalent to a $\chi^2$ minimisation for a straight line.

The straight line is parameterised in the xy and the zy plane:
x=ay+b ; z=c*y+d
Calculate a and b ; c and d analytically by linear regression. (As track is going predominantly along the y direction, this coordinate is set to the centre of the pad and regarded without uncertainty)

$a=\frac{\sum y_i\sum x_i-n\sum y_i x_i} {(\sum y_i)^2 - n \sum y_i^2}$