RANSACModel< MODEL, DATA > Class Template Reference

An interface for RANSAC model fitting. More...

#include <RANSACModel.hpp>

Public Types
typedef rw::core::Ptr< RANSACModel< MODEL, DATA > >	Ptr
	Smart pointer type to this class.

Public Member Functions
	RANSACModel ()
	Constructor.
virtual MODEL	make (const std::vector< DATA > &data) const
	Creates a new model of this type using provided data.
virtual	~RANSACModel ()
	Destructor.
bool	operator< (const MODEL &model) const
	'Worse than' operator. More...
bool	operator> (const MODEL &model) const
	'Better than' operator. More...
virtual double	fitError (const DATA &sample) const =0
	Calculates the fitting error of a sample.
virtual bool	belongsTo (const DATA &sample, double threshold) const
	Check whether a sample belongs to the model. More...
virtual bool	invalid () const =0
	Checks whether the model is invalid.
virtual double	refit (const std::vector< DATA > &data)=0
	Recalculates the model based on provided samples. More...
virtual int	getMinReqData () const =0
	Returns the number of samples required to create the model. More...
virtual bool	same (const MODEL &model, double threshold) const =0
	Tests whether the model is same to a threshold of another model. More...
size_t	getNumberOfInliers () const
	Get the number of inliers.
double	getQuality () const
	Get the model quality. More...
void	setQuality (double quality)
	Set the model quality. More...
std::vector< DATA > &	getData ()
	Access data.
const std::vector< DATA > &	getData () const
	Access data.
std::vector< size_t >	getInlierIndices () const
	Get the vector of inlier indices. More...

Static Public Member Functions
static std::vector< MODEL >	findModels (const std::vector< DATA > &data, int maxIterations, int dataRequired, double dataThreshold, double modelThreshold)
	Find models fitting a set of observations. More...
static MODEL	bestModel (const std::vector< MODEL > &models)
	Select the model with the largest number of inliers. More...
static MODEL	likelyModel (const std::vector< MODEL > &models)
	Select a model randomly, with a chance based on the number of inliers. More...

Protected Attributes
std::vector< DATA >	_data
double	_quality
std::vector< size_t >	_indices

Detailed Description

template<class MODEL, class DATA>
class rwlibs::algorithms::RANSACModel< MODEL, DATA >

An interface for RANSAC model fitting.

Todo:

A model needs to remember the indices of inliers from the set of data...

Member Function Documentation

belongsTo()

virtual bool belongsTo ( const DATA &  sample, double  threshold  ) const

inlinevirtual

Check whether a sample belongs to the model.

Returns true when the sample is within a threshold distance of the model.

bestModel()

static MODEL bestModel ( const std::vector< MODEL > &  models )

inlinestatic

Select the model with the largest number of inliers.

In case of ties, pick the model with better quality.

findModels()

static std::vector<MODEL> findModels ( const std::vector< DATA > &  data, int  maxIterations, int  dataRequired, double  dataThreshold, double  modelThreshold  )

inlinestatic

Find models fitting a set of observations.

Function performs maxIterations iterations and finds a number of initial models. The models are then filtered, and those within a distance of modelThreshold are merged together.

Parameters

data	[in] set of data points to find a model for
maxIterations	[in] number of iterations to perform
dataRequired	[in] a number of data points required to fit into model to consider it plausible
dataThreshold	[in] a threshold for fitting error of a data point
modelThreshold	[in] a difference between models neccesary to consider them different

Returns

a vector of fitted models

getInlierIndices()

std::vector<size_t> getInlierIndices ( ) const

inline

Get the vector of inlier indices.

Returns

a vector of indices of inliers in the data vector supplied to findModels function.

getMinReqData()

virtual int getMinReqData ( ) const

pure virtual

Returns the number of samples required to create the model.

No model will be found, if data size is less than this amount.

Implemented in StructuredLineModel, StablePose1DModel, StablePose0DModel, PointModel, PlaneModel, and LineModel.

getQuality()

double getQuality ( ) const

inline

Get the model quality.

The model quality is a sum of fitting errors for all its inliers.

likelyModel()

static MODEL likelyModel ( const std::vector< MODEL > &  models )

inlinestatic

Select a model randomly, with a chance based on the number of inliers.

Given a vector of models {model1(45 inliers), model2(30 inliers), model3(20 inliers), model4(5 inliers)}, it will pick:

• model1 - 45% chance,
• model2 - 30% chance,
• model3 - 20% chance,
• model4 - 5% chance

operator<()

bool operator< ( const MODEL &  model ) const

inline

'Worse than' operator.

Used for sorting. Compares the number of inliers the models have. In case of ties, compares quality.

operator>()

bool operator> ( const MODEL &  model ) const

inline

'Better than' operator.

Used for sorting. Compares the number of inliers the models have. In case of ties, compares quality.

refit()

virtual double refit ( const std::vector< DATA > &  data )

pure virtual

Recalculates the model based on provided samples.

Returns

Fit error on a set of provided samples.

same()

virtual bool same ( const MODEL &  model, double  threshold  ) const

pure virtual

Tests whether the model is same to a threshold of another model.

This is used to filter down (merge) similar models found in the course of RANSAC algorithm execution.

Implemented in StructuredLineModel, StablePose1DModel, StablePose0DModel, PointModel, PlaneModel, and LineModel.

setQuality()

void setQuality ( double  quality )

inline

Set the model quality.

Parameters

quality

[in] fitting error of the model

---

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

• RANSACModel.hpp