org.ddogleg.fitting.modelset.distance

Class StatisticalDistanceModelMatcher<Model,Point>

java.lang.Object

org.ddogleg.fitting.modelset.distance.StatisticalDistanceModelMatcher<Model,Point>

All Implemented Interfaces:

ModelMatcher<Model,Point>

public class StatisticalDistanceModelMatcher<Model,Point>
extends java.lang.Object
implements ModelMatcher<Model,Point>

Outliers are removed by first fitting a model to all the data points. Points which deviate the most are removed and the best fit parameters are recomputed. This is done until some error metric changes very little or the maximum number of iterations has been exceeded. Works well when the inlier set is much greater than the noise.

Anyone have a good reference paper for this approach?

Constructor Summary

Constructors

Constructor and Description
StatisticalDistanceModelMatcher(int maxIterations, double minChange, double exitCenterError, double failError, int minFitPoints, StatisticalDistance statistics, double pruneThreshold, ModelManager<Model> modelManager, ModelFitter<Model,Point> modelFitter, DistanceFromModel<Model,Point> modelError, ModelCodec<Model> codec) Creates a new model matcher.

Method Summary

Modifier and Type	Method and Description
double	getFitQuality() Returns the metric used to evaluate the quality of fit.
int	getInputIndex(int matchIndex) For an item in the match set, return the index of the item in the original input set.
java.util.List<Point>	getMatchSet() A set of points which match the provided parameters.
int	getMinimumSize() This is the minimum number of observations which can be input and produce a valid model.
Model	getModelParameters() Model for the match set
boolean	process(java.util.List<Point> dataSet) Finds a set of points from the provided list that are a good fit for the internal model and computes the fit parameters for the model.

Methods inherited from class java.lang.Object

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

Constructor Detail

StatisticalDistanceModelMatcher

public StatisticalDistanceModelMatcher(int maxIterations,
                                       double minChange,
                                       double exitCenterError,
                                       double failError,
                                       int minFitPoints,
                                       StatisticalDistance statistics,
                                       double pruneThreshold,
                                       ModelManager<Model> modelManager,
                                       ModelFitter<Model,Point> modelFitter,
                                       DistanceFromModel<Model,Point> modelError,
                                       ModelCodec<Model> codec)

Creates a new model matcher. The type of statistics it uses is specified by "statistics". 0 for mean and standard deviation and 1 for median and percentile. If statistics=0 is set then the threshold correspond to the number of standard deviations a point can be away from the mean before it is pruned. If statistics=1 then points which have a percentile error more than that value are pruned.

Parameters:

maxIterations - The maximum number of iterations it will perform.

minChange - It will stop iterating of the change in error is less than this amount.

exitCenterError - If the error is less than this value it will stop iterating.

failError - If the final error is more than this amount it failed.

minFitPoints - If fewer than this number of points remain, then it failed.

statistics - 0 = mean statistics and 1 = percentile statistics

pruneThreshold - Points which exceed this statistic are pruned. See StatisticalDistance for details.

modelFitter - Fits a model to a set of points

modelError - Computes the error between a point and the model.

Method Detail

process

public boolean process(java.util.List<Point> dataSet)

Description copied from interface: ModelMatcher

Finds a set of points from the provided list that are a good fit for the internal model and computes the fit parameters for the model.

Specified by:

process in interface ModelMatcher<Model,Point>

Parameters:

dataSet - Set of points (with noise) that are to be fit.

Returns:

true if it successfully found a solution or false if not.

getModelParameters

public Model getModelParameters()

Description copied from interface: ModelMatcher

Model for the match set

Specified by:

getModelParameters in interface ModelMatcher<Model,Point>

Returns:

model.

getMatchSet

public java.util.List<Point> getMatchSet()

Description copied from interface: ModelMatcher

A set of points which match the provided parameters.

Specified by:

getMatchSet in interface ModelMatcher<Model,Point>

Returns:

List of points in the match set.

getInputIndex

public int getInputIndex(int matchIndex)

Description copied from interface: ModelMatcher

For an item in the match set, return the index of the item in the original input set.

Specified by:

getInputIndex in interface ModelMatcher<Model,Point>

Parameters:

matchIndex - Index of an element in the match set.

Returns:

Index of the same element in the original input list.

getFitQuality

public double getFitQuality()

Description copied from interface: ModelMatcher

Returns the metric used to evaluate the quality of fit. Meaning is implementation specific. Larger or smaller values could be preferred depending on implementation..

Specified by:

getFitQuality in interface ModelMatcher<Model,Point>

Returns:

Quality of fit to matched set of points

getMinimumSize

public int getMinimumSize()

Description copied from interface: ModelMatcher

This is the minimum number of observations which can be input and produce a valid model.

Specified by:

getMinimumSize in interface ModelMatcher<Model,Point>

Returns:

Minimum number of sample points