ISOData strategy for an unsupervised pixel-based classification algorithm. More...

#include <ISOData.h>

Classes
class	ClusterData
	Cluster data. More...

class	Parameters
	Classifier Parameters. More...

Public Member Functions
bool	classify (const InputAdaptor< double > &input, const std::vector< unsigned int > &attributesIndices, const std::vector< double > &inputNoDataValues, InputOutputAdaptor< ClusterData::ClusterLabelT > &output, const unsigned int outputIndex, const ClusterData::ClusterLabelT outputNoDataValue, const bool enableProgressInterface, std::vector< ClusterData > *clustersDataPtr)
	Classify an input iterated data and save the result on the output iterated data.

bool	initialize (const Parameters &params)
	Initialize this classifier instance with new parameters.

	ISOData ()

	~ISOData ()

Protected Types
typedef std::map< ClusterData::ClusterLabelT, ISOData::ClusterData >	ClustersContainerT

Protected Member Functions
bool	classifyElements (const InputAdaptor< double > &input, const std::vector< unsigned int > &attributesIndices, const std::vector< double > &inputNoDataValues, ClustersContainerT &clustersMap, InputOutputAdaptor< unsigned int > &output, const unsigned int outputIndex, const unsigned int outputNoDataValue, const std::vector< std::pair< double, double > > &inputElementsOffsetsAndGains) const
	Classify each element following clusers centers.

bool	initializeClusters (const InputAdaptor< double > &input, const std::vector< unsigned int > &attributesIndices, ClustersContainerT &clustersMap, ClusterData::ClusterLabelT &nextAvaliableClusterLabel, const std::vector< double > &inputNoDataValues, std::vector< std::pair< double, double > > &inputElementsOffsetsAndGains) const
	Initialize clusters.

bool	removeSmallClusters (const unsigned int inputElelementsCount, ClustersContainerT &clustersMap) const
	Remove too mall clusters following the current parameters.

void	reset ()
	Reset this instance to its initial state.

bool	splitClusters (const unsigned int inputElementsCount, const unsigned int maxNumberOfSplits, ClusterData::ClusterLabelT &nextAvaliableClusterLabel, ClustersContainerT &clustersMap) const
	Split clusters.

bool	updateClustersStats (const InputAdaptor< double > &input, const std::vector< unsigned int > &attributesIndices, const std::vector< double > &inputNoDataValues, ClustersContainerT &clustersMap, const InputOutputAdaptor< unsigned int > &output, const unsigned int outputIndex, const std::vector< std::pair< double, double > > &inputElementsOffsetsAndGains) const
	Update clusters statistical data.

Static Protected Member Functions
static bool	mergeClusters (const double mergeDistThreshold, const unsigned int maxNumberOfMerges, ClustersContainerT &clustersMap)
	Merge clusters.

static void	printClusters (const ClustersContainerT &clustersMap)

static void	recodeClustersLabels (ClustersContainerT &clustersMap)
	Recode clusters labels to be in range zero up to the number of clusters - 1.

Protected Attributes
bool	m_isInitialized
	True if this instance is initialized.

Parameters	m_parameters
	Internal execution parameters.

Detailed Description

ISOData strategy for an unsupervised pixel-based classification algorithm.

Note: Based on Reference: IsoData Algorithm - Pattern Recognition Principles - Rafael Gonzalez, Julius T. Tou - Addison Wesley Publishing Company - 1974.

Definition at line 51 of file ISOData.h.

Member Typedef Documentation

◆ ClustersContainerT

typedef std::map< ClusterData::ClusterLabelT, ISOData::ClusterData > te::cl::ISOData::ClustersContainerT

protected

Definition at line 144 of file ISOData.h.

Constructor & Destructor Documentation

◆ ISOData()

te::cl::ISOData::ISOData ( )

◆ ~ISOData()

te::cl::ISOData::~ISOData ( )

Member Function Documentation

◆ classify()

bool te::cl::ISOData::classify	(	const InputAdaptor< double > &	input,
		const std::vector< unsigned int > &	attributesIndices,
		const std::vector< double > &	inputNoDataValues,
		InputOutputAdaptor< ClusterData::ClusterLabelT > &	output,
		const unsigned int	outputIndex,
		const ClusterData::ClusterLabelT	outputNoDataValue,
		const bool	enableProgressInterface,
		std::vector< ClusterData > *	clustersDataPtr
	)

Classify an input iterated data and save the result on the output iterated data.

Parameters

input	Input data to be classified.
attributesIndices	The attributes indexes to process from the iterated train data.
inputNoDataValues	A vector of no-data values for each attribute dimension or an empty vector if no-data values are not used.
output	Output classified data.
outputIndex	The output attribute index.
outputNoDataValue	A output label value to use when dealing with input no-data.
enableProgressInterface	Enable/disable the use of a progress interface.
clustersDataPtr	A pointer to a vector where the current clusters data will be storer or a null pointer.

◆ classifyElements()

bool te::cl::ISOData::classifyElements	(	const InputAdaptor< double > &	input,
		const std::vector< unsigned int > &	attributesIndices,
		const std::vector< double > &	inputNoDataValues,
		ClustersContainerT &	clustersMap,
		InputOutputAdaptor< unsigned int > &	output,
		const unsigned int	outputIndex,
		const unsigned int	outputNoDataValue,
		const std::vector< std::pair< double, double > > &	inputElementsOffsetsAndGains
	)		const

protected

Classify each element following clusers centers.

Parameters

input	Input data to be classified.
attributesIndices	The attributes indexes to process from the iterated train data.
inputNoDataValues	A vector of no-data values for each attribute dimension or an empty vector if no-data values are not used.
output	Output classified data.
outputIndex	The output attribute index.
clustersMap	Clusters map.
outputNoDataValue	Output no-data value.
inputElementsOffsetsAndGains	Elements offsets (first) and gains (second).

Note: Only the member ClusterData::m_means is used for this operation.; Valid updated members after: ClusterData::m_label, ClusterData::m_means, ClusterData::m_sums, ClusterData::m_size

◆ initialize()

bool te::cl::ISOData::initialize ( const Parameters & params )

Initialize this classifier instance with new parameters.

Parameters

params New initialization parameters.

◆ initializeClusters()

bool te::cl::ISOData::initializeClusters	(	const InputAdaptor< double > &	input,
		const std::vector< unsigned int > &	attributesIndices,
		ClustersContainerT &	clustersMap,
		ClusterData::ClusterLabelT &	nextAvaliableClusterLabel,
		const std::vector< double > &	inputNoDataValues,
		std::vector< std::pair< double, double > > &	inputElementsOffsetsAndGains
	)		const

protected

Initialize clusters.

Parameters

input	Input data to be classified.
attributesIndices	The attributes indexes to process from the iterated train data.
inputNoDataValues	A vector of no-data values for each attribute dimension or an empty vector if no-data values are not used.
clustersMap	Clusters map.
nextAvaliableClusterLabel	Reference to the global clusters labels counter.
inputElementsOffsetsAndGains	Elements offsets (first) and gains (second).

Note: Updated clusters members: ClusterData::m_means, ClusterData::m_label.; Output = ( input + offset ) * gain;

◆ mergeClusters()

static bool te::cl::ISOData::mergeClusters	(	const double	mergeDistThreshold,
		const unsigned int	maxNumberOfMerges,
		ClustersContainerT &	clustersMap
	)

staticprotected

Merge clusters.

Parameters

clustersMap	Clusters map.
mergeDistThreshold	A distance treshold to control the merge process (higher values will allow the merge more clusters. Valid values: zero or positive values ).
maxNumberOfMerges	Maximum number of merges.

Note: Only the members ClusterData::m_label, ClusterData::m_means, ClusterData::m_size are used for this operation.; Only the following members will be valid after merging: ClusterData::m_means, ClusterData::m_label, ClusterData::m_size.

◆ printClusters()

static void te::cl::ISOData::printClusters ( const ClustersContainerT & clustersMap )

staticprotected

◆ recodeClustersLabels()

static void te::cl::ISOData::recodeClustersLabels ( ClustersContainerT & clustersMap )

staticprotected

Recode clusters labels to be in range zero up to the number of clusters - 1.

Parameters

clustersMap Clusters map.

◆ removeSmallClusters()

bool te::cl::ISOData::removeSmallClusters	(	const unsigned int	inputElelementsCount,
		ClustersContainerT &	clustersMap
	)		const

protected

Remove too mall clusters following the current parameters.

Parameters

inputElelementsCount	Input data elements counte.
clustersMap	Clusters map.

Note: Only the member ClusterData::m_size is used for this operation.

◆ reset()

void te::cl::ISOData::reset ( )

protected

Reset this instance to its initial state.

◆ splitClusters()

bool te::cl::ISOData::splitClusters	(	const unsigned int	inputElementsCount,
		const unsigned int	maxNumberOfSplits,
		ClusterData::ClusterLabelT &	nextAvaliableClusterLabel,
		ClustersContainerT &	clustersMap
	)		const

protected

Split clusters.

Parameters

clustersMap	Clusters map.
inputElementsCount	Input data total elements number.
maxNumberOfSplits	Maximum number of splits.
nextAvaliableClusterLabel	Reference to the global clusters labels counter.

Note: Only the members ClusterData::m_label, ClusterData::m_means, ClusterData::m_stdDevs, ClusterData::m_size, ClusterData::m_avgDistToClusterCenter are used for this operation.; Only the following members will be valid after splitting: ClusterData::m_means, ClusterData::m_label.

◆ updateClustersStats()

bool te::cl::ISOData::updateClustersStats	(	const InputAdaptor< double > &	input,
		const std::vector< unsigned int > &	attributesIndices,
		const std::vector< double > &	inputNoDataValues,
		ClustersContainerT &	clustersMap,
		const InputOutputAdaptor< unsigned int > &	output,
		const unsigned int	outputIndex,
		const std::vector< std::pair< double, double > > &	inputElementsOffsetsAndGains
	)		const

protected

Update clusters statistical data.

Parameters

input	Input data to be classified.
attributesIndices	The attributes indexes to process from the iterated train data.
inputNoDataValues	A vector of no-data values for each attribute dimension or an empty vector if no-data values are not used.
output	Output classified data.
outputIndex	The output attribute index.
clustersMap	Clusters map.
inputElementsOffsetsAndGains	Elements offsets (first) and gains (second).

Note: Used clusters members: ClusterData::m_label, ClusterData::m_means, ClusterData::m_size; Updated clusters members: ClusterData::m_avgDistToClusterCenter, ClusterData::m_stdDevs.

Member Data Documentation

◆ m_isInitialized

bool te::cl::ISOData::m_isInitialized

protected

True if this instance is initialized.

Definition at line 146 of file ISOData.h.

◆ m_parameters

Parameters te::cl::ISOData::m_parameters

protected

Internal execution parameters.

Definition at line 147 of file ISOData.h.

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

/home/castejon/castejon_files/develop/terralib5/git_release/src/terralib/classification/ISOData.h

Classes

Public Member Functions

Protected Types

Protected Member Functions

Static Protected Member Functions

Protected Attributes

Detailed Description

Member Typedef Documentation

◆ ClustersContainerT

Constructor & Destructor Documentation

◆ ISOData()

◆ ~ISOData()

Member Function Documentation

◆ classify()

◆ classifyElements()

◆ initialize()

◆ initializeClusters()

◆ mergeClusters()

◆ printClusters()

◆ recodeClustersLabels()

◆ removeSmallClusters()

◆ reset()

◆ splitClusters()

◆ updateClustersStats()

Member Data Documentation

◆ m_isInitialized

◆ m_parameters