Package net.preibisch.mvrecon.process.pointcloud.icp

Class ICP<P extends net.imglib2.RealLocalizable>

java.lang.Object

net.preibisch.mvrecon.process.pointcloud.icp.ICP<P>

Type Parameters:

P - - something extening point

public class ICP<P extends net.imglib2.RealLocalizable>
extends Object

Implementation of the ICP, takes anything RealLocalizable and internally uses LinkedInterestPoint to compute the ICP

Author:

Stephan Preibisch

Constructor Summary

Constructors

Constructor	Description
ICP(List<P> target, List<P> reference, boolean useRANSAC, double minInlierRatio, double maxEpsilonRANSAC, int maxIterationsRANSAC)	Also instantiates a new ICP instance, but uses the SimplePointMatchIdentification to define corresponding points.
ICP(List<P> target, List<P> reference, double distanceThreshold, boolean useRANSAC, double minInlierRatio, double maxEpsilonRANSAC, int maxIterationsRANSAC)	Also instantiates a new ICP instance, but uses the SimplePointMatchIdentification to define corresponding points.
ICP(List<P> target, List<P> reference, PointMatchIdentification<P> pointMatchIdentifier, boolean useRANSAC, double minInlierRatio, double maxEpsilonRANSAC, int maxIterationsRANSAC)	Instantiates a new ICP object with the List of target and reference points as well as the PointMatchIdentification interface that defines how corresponding points are identified.

Method Summary

Modifier and Type	Method	Description
void	estimateIntialModel(List<PointMatchGeneric<LinkedInterestPoint<P>>> matches, mpicbg.models.Model<?> model)	Estimates an initial Model based on some given PointMatches.
ArrayList<mpicbg.models.PointMatch>	getAmbigousMatches()	Returns the ArrayList of ambigous PointMatches indentified in the last ICP iteration, or null if no iteration has been computed yet.
double	getAverageError()	Returns the average error of the last ICP iteration, or -1 if no iteration has been computed yet.
double	getMaximalError()	Returns the maximum error of a PointMatch of the last ICP iteration, or -1 if no iteration has been computed yet.
int	getNumAmbigousMatches()	Returns the number of ambigous PointMatches indentified in the last ICP iteration, or -1 if no iteration has been computed yet.
int	getNumPointMatches()	Returns the number of PointMatches of the last ICP iteration, or -1 if no iteration has been computed yet.
protected static <P extends mpicbg.models.Point & net.imglib2.RealLocalizable> ArrayList<Integer>	getOccurences(mpicbg.models.Point pointTarget, mpicbg.models.Point pointReference, List<PointMatchGeneric<P>> list)	Computes if one Point pair occurs more than once in a List of PointMatches
List<PointMatchGeneric<LinkedInterestPoint<P>>>	getPointMatches()	Return the List of PointMatches (target, reference) of the last ICP iteration
PointMatchIdentification<P>	getPointMatchIdentification()	Returns the current PointMatchIdentification that is used to identfy corresponding points
List<LinkedInterestPoint<P>>	getWrappedReferencePoints()
List<LinkedInterestPoint<P>>	getWrappedTargetPoints()
static double	maxDistance(Collection<? extends mpicbg.models.PointMatch> matches)
static double	meanDistance(Collection<? extends mpicbg.models.PointMatch> matches)
static <P extends mpicbg.models.Point & net.imglib2.RealLocalizable> ArrayList<mpicbg.models.PointMatch>	removeAmbigousMatches(List<PointMatchGeneric<P>> matches)	Detects ambigous (and duplicate) PointMatches, i.e. if a Point corresponds with more than one other Point
void	runICPIteration(mpicbg.models.Model<?> lastModel, mpicbg.models.Model<?> newModel)	Performs one iteration of the ICP.
void	setPointMatchIdentification(PointMatchIdentification<P> pointMatchIdentifier)	Sets the PointMatchIdentification that defines how PointMatches between reference and target are identified.
static <P extends mpicbg.models.Point> List<PointMatchGeneric<P>>	unwrapPointMatches(List<? extends PointMatchGeneric<? extends LinkedInterestPoint<P>>> pointMatches)	Return the List of PointMatches (target, reference) of the last ICP iteration

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

ICP

public ICP(List<P> target,
 List<P> reference,
 PointMatchIdentification<P> pointMatchIdentifier,
 boolean useRANSAC,
 double minInlierRatio,
 double maxEpsilonRANSAC,
 int maxIterationsRANSAC)

Instantiates a new ICP object with the List of target and reference points as well as the PointMatchIdentification interface that defines how corresponding points are identified.

Parameters:

target - - the List of target points

reference - - the List of reference points

pointMatchIdentifier - - the PointMatchIdentification which defines how correspondences are established

useRANSAC - - use RANSAC during every iteration

minInlierRatio - - minInlierRatio for RANSAC

maxEpsilonRANSAC - - max error for RANSAC if used

maxIterationsRANSAC - - max iterations for RANSAC if used

ICP

public ICP(List<P> target,
 List<P> reference,
 double distanceThreshold,
 boolean useRANSAC,
 double minInlierRatio,
 double maxEpsilonRANSAC,
 int maxIterationsRANSAC)

Also instantiates a new ICP instance, but uses the SimplePointMatchIdentification to define corresponding points.

Parameters:

target - - the List of target points

reference - - the List of reference points

distanceThreshold - - the maximal distance of SimplePointMatchIdentification, so that the nearest neighbor of a point is still counted as a corresponding point

useRANSAC - - use RANSAC during every iteration

minInlierRatio - - minInlierRatio for RANSAC

maxEpsilonRANSAC - - max error for RANSAC if used

maxIterationsRANSAC - - max iterations for RANSAC if used

ICP

public ICP(List<P> target,
 List<P> reference,
 boolean useRANSAC,
 double minInlierRatio,
 double maxEpsilonRANSAC,
 int maxIterationsRANSAC)

Also instantiates a new ICP instance, but uses the SimplePointMatchIdentification to define corresponding points.

Parameters:

target - - the List of target points

reference - - the List of reference points

useRANSAC - - use RANSAC during every iteration

minInlierRatio - - minInlierRatio for RANSAC

maxEpsilonRANSAC - - max error for RANSAC if used

maxIterationsRANSAC - - max iterations for RANSAC if used

Method Details

runICPIteration

public void runICPIteration(mpicbg.models.Model<?> lastModel,
 mpicbg.models.Model<?> newModel)
                     throws mpicbg.models.NotEnoughDataPointsException,
mpicbg.models.IllDefinedDataPointsException,
NoSuitablePointsException

Performs one iteration of the ICP. It takes the last Model as input to find the corresponding points for the new Model. The result is the new Model, the number of corresponding points, the average, and the maximal error. Note that lastModel and newModel can be the same instance and it will be overwritten.

Parameters:

lastModel - - The last Model that maps the target.local coordinates to the reference.world coordinates, used to find the corresponding points

newModel - - The Model that maps the target.local coordinates to the reference.world coordinates, will be fitted to the new points

Throws:

mpicbg.models.NotEnoughDataPointsException - - if there are not enough points

mpicbg.models.IllDefinedDataPointsException - - if the points are ill-defined

NoSuitablePointsException - - if points are somehow wrong

estimateIntialModel

public void estimateIntialModel(List<PointMatchGeneric<LinkedInterestPoint<P>>> matches,
 mpicbg.models.Model<?> model)
                         throws mpicbg.models.NotEnoughDataPointsException,
mpicbg.models.IllDefinedDataPointsException

Estimates an initial Model based on some given PointMatches. Note that the PointMatches have to be stored as PointMatch(target,reference).

Parameters:

matches - - The List of apriori known PointMatches

model - - The Model to use

Throws:

mpicbg.models.NotEnoughDataPointsException - - if there are not enough points

mpicbg.models.IllDefinedDataPointsException - - if the points are ill-defined

unwrapPointMatches

public static <P extends mpicbg.models.Point>
List<PointMatchGeneric<P>> unwrapPointMatches(List<? extends PointMatchGeneric<? extends LinkedInterestPoint<P>>> pointMatches)

Return the List of PointMatches (target, reference) of the last ICP iteration

Type Parameters:

P - - something that extens point

Parameters:

pointMatches - - the matches

Returns:

- List of PointMatches

setPointMatchIdentification

public void setPointMatchIdentification(PointMatchIdentification<P> pointMatchIdentifier)

Sets the PointMatchIdentification that defines how PointMatches between reference and target are identified. The simplest way to do it is the SimplePointMatchIdentification class which takes the nearest neighbor with a minimal distance threshold.

Parameters:

pointMatchIdentifier - - the new PointMatchIdentification

getPointMatchIdentification

public PointMatchIdentification<P> getPointMatchIdentification()

Returns the current PointMatchIdentification that is used to identfy corresponding points

Returns:

PointMatchIdentification> P <

getPointMatches

public List<PointMatchGeneric<LinkedInterestPoint<P>>> getPointMatches()

Return the List of PointMatches (target, reference) of the last ICP iteration

Returns:

- List of PointMatches

getAverageError

public double getAverageError()

Returns the average error of the last ICP iteration, or -1 if no iteration has been computed yet.

Returns:

double - average error

getMaximalError

public double getMaximalError()

Returns the maximum error of a PointMatch of the last ICP iteration, or -1 if no iteration has been computed yet.

Returns:

double - maximal error

getNumPointMatches

public int getNumPointMatches()

Returns the number of PointMatches of the last ICP iteration, or -1 if no iteration has been computed yet.

Returns:

int - number of PointMatches

getNumAmbigousMatches

public int getNumAmbigousMatches()

Returns the number of ambigous PointMatches indentified in the last ICP iteration, or -1 if no iteration has been computed yet.

Returns:

int - number of ambigous PointMatches

getAmbigousMatches

public ArrayList<mpicbg.models.PointMatch> getAmbigousMatches()

Returns the ArrayList of ambigous PointMatches indentified in the last ICP iteration, or null if no iteration has been computed yet.

Returns:

int - ArrayList of PointMatches

getWrappedTargetPoints

public List<LinkedInterestPoint<P>> getWrappedTargetPoints()

getWrappedReferencePoints

public List<LinkedInterestPoint<P>> getWrappedReferencePoints()

removeAmbigousMatches

public static <P extends mpicbg.models.Point & net.imglib2.RealLocalizable>
ArrayList<mpicbg.models.PointMatch> removeAmbigousMatches(List<PointMatchGeneric<P>> matches)

Detects ambigous (and duplicate) PointMatches, i.e. if a Point corresponds with more than one other Point

Type Parameters:

P - - something extening point

Parameters:

matches - - the List of PointMatches

Returns:

- the ArrayList containing the removed ambigous or duplicate PointMatches

getOccurences

protected static <P extends mpicbg.models.Point & net.imglib2.RealLocalizable>
ArrayList<Integer> getOccurences(mpicbg.models.Point pointTarget,
 mpicbg.models.Point pointReference,
 List<PointMatchGeneric<P>> list)

Computes if one Point pair occurs more than once in a List of PointMatches

Type Parameters:

P - - something extening point

Parameters:

pointTarget - - one Point

pointReference - - the other Point

list - - the List of PointMatches (target, reference)

Returns:

- an ArrayList of indices which should be removed due to duplicate or ambigous occurence

meanDistance

public static double meanDistance(Collection<? extends mpicbg.models.PointMatch> matches)

maxDistance

public static double maxDistance(Collection<? extends mpicbg.models.PointMatch> matches)