org.openimaj.math.geometry.transforms.estimation

Class RobustHomographyEstimator

java.lang.Object

org.openimaj.math.geometry.transforms.estimation.RobustHomographyEstimator

All Implemented Interfaces:

ModelFitting<Point2d,Point2d,HomographyModel>, RobustModelFitting<Point2d,Point2d,HomographyModel>

public class RobustHomographyEstimator
extends Object
implements RobustModelFitting<Point2d,Point2d,HomographyModel>

Helper class to simplify robust estimation of homographies without having to deal with the nuts and bolts of the underlying robust model fitters, etc. The overall robust estimation process is as follows:

An initial estimate of the inliers and an algebraically optimal homography is computed using RANSAC or LMedS with a BucketingSampler2d sampling strategy for selecting subsets. In both cases, the normalised DLT algorithm is used (see TransformUtilities.homographyMatrixNorm(List)

If an reasonable initial estimate was found, non-linear optimisation using Levenburg-Marquardt is performed to on the inliers using the initial estimate to optimise against a true geometric residual given by HomographyRefinement.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
RobustHomographyEstimator(double outlierProportion, HomographyRefinement refinement) Construct using the LMedS algorithm with the given expected outlier percentage
RobustHomographyEstimator(double outlierProportion, HomographyRefinement refinement, Predicate<HomographyModel> modelCheck) Construct using the LMedS algorithm with the given expected outlier percentage
RobustHomographyEstimator(double threshold, int nIterations, RANSAC.StoppingCondition stoppingCondition, HomographyRefinement refinement) Construct using the RANSAC algorithm with the given options.
RobustHomographyEstimator(double threshold, int nIterations, RANSAC.StoppingCondition stoppingCondition, HomographyRefinement refinement, Predicate<HomographyModel> modelCheck) Construct using the RANSAC algorithm with the given options.

Method Summary

Modifier and Type	Method and Description
boolean	fitData(List<? extends IndependentPair<Point2d,Point2d>> data) Attempt to fit the given data to the model.
List<? extends IndependentPair<Point2d,Point2d>>	getInliers()
HomographyModel	getModel()
List<? extends IndependentPair<Point2d,Point2d>>	getOutliers()
int	numItemsToEstimate()

Methods inherited from class java.lang.Object

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

Constructor Detail

RobustHomographyEstimator

public RobustHomographyEstimator(double outlierProportion,
                                 HomographyRefinement refinement)

Construct using the LMedS algorithm with the given expected outlier percentage

Parameters:

outlierProportion - expected proportion of outliers (between 0 and 1)

refinement - the refinement technique

RobustHomographyEstimator

public RobustHomographyEstimator(double threshold,
                                 int nIterations,
                                 RANSAC.StoppingCondition stoppingCondition,
                                 HomographyRefinement refinement)

Construct using the RANSAC algorithm with the given options.

Parameters:

threshold - the threshold on the AlgebraicResidual2d at which to consider a point as an inlier

nIterations - the maximum number of iterations

stoppingCondition - the RANSAC.StoppingCondition for RANSAC

refinement - the refinement technique

RobustHomographyEstimator

public RobustHomographyEstimator(double outlierProportion,
                                 HomographyRefinement refinement,
                                 Predicate<HomographyModel> modelCheck)

Construct using the LMedS algorithm with the given expected outlier percentage

Parameters:

outlierProportion - expected proportion of outliers (between 0 and 1)

refinement - the refinement technique

modelCheck - the predicate to test whether an estimated model is sane

RobustHomographyEstimator

public RobustHomographyEstimator(double threshold,
                                 int nIterations,
                                 RANSAC.StoppingCondition stoppingCondition,
                                 HomographyRefinement refinement,
                                 Predicate<HomographyModel> modelCheck)

Construct using the RANSAC algorithm with the given options.

Parameters:

threshold - the threshold on the AlgebraicResidual2d at which to consider a point as an inlier

nIterations - the maximum number of iterations

stoppingCondition - the RANSAC.StoppingCondition for RANSAC

refinement - the refinement technique

modelCheck - the predicate to test whether an estimated model is sane

Method Detail

fitData

public boolean fitData(List<? extends IndependentPair<Point2d,Point2d>> data)

Description copied from interface: ModelFitting

Attempt to fit the given data to the model.

Specified by:

fitData in interface ModelFitting<Point2d,Point2d,HomographyModel>

Parameters:

data - Data to be fitted

Returns:

true on success, false otherwise

numItemsToEstimate

public int numItemsToEstimate()

Specified by:

numItemsToEstimate in interface ModelFitting<Point2d,Point2d,HomographyModel>

Returns:

The minimum number of observations required to estimate the model.

getModel

public HomographyModel getModel()

Specified by:

getModel in interface ModelFitting<Point2d,Point2d,HomographyModel>

Returns:

the trained model object

getInliers

public List<? extends IndependentPair<Point2d,Point2d>> getInliers()

Specified by:

getInliers in interface RobustModelFitting<Point2d,Point2d,HomographyModel>

Returns:

list of the inliers in the original data

getOutliers

public List<? extends IndependentPair<Point2d,Point2d>> getOutliers()

Specified by:

getOutliers in interface RobustModelFitting<Point2d,Point2d,HomographyModel>

Returns:

list of the outliers in the original data