KMeans strategy for image classification. Step-by-step: More...

#include <ClassifierKMeansStrategy.h>

Inheritance diagram for te::rp::ClassifierKMeansStrategy:

Classes
class	Parameters
	Classifier Parameters. More...

Public Member Functions
	ClassifierKMeansStrategy ()

void	enableMultiThread (const bool &enable)
	Enable / disable the use of multiple threads. More...

void	enableOutputPalette (const bool enabled)
	Enable (true) or disable (false) the creation of a paletted output raster. More...

void	enableProgressInterface (const bool &enable)
	Enable / disable the progress interface. More...

void	enableRasterCache (const bool &enable)
	Enable / disable the use of raster data cache. More...

bool	execute () throw (te::rp::Exception)
	Executes the classification strategy. More...

const std::string &	getErrorMessage () const
	Return the current error message if there is any. More...

void	getMetaData (std::map< std::string, std::string > &metadata) const
	Returns strategy-dependent metadata. More...

bool	initialize (ClassifierStrategyParameters const *const strategyParams) throw (te::rp::Exception)
	Initialize the classification strategy. More...

std::unique_ptr< te::rst::Raster >	releaseOutputRaster ()
	Returns a pointer to the output raster or a void pointer if there is none. More...

void	setInputPolygons (const std::vector< te::gm::Polygon * > &polygonsPtrs)
	Set the input polygons. More...

void	setInputRaster (const te::rst::Raster &raster)
	Set the input raster. More...

void	setInputRasterBands (const std::vector< unsigned int > &rasterBands)
	Set the input raste bandsr. More...

void	setUserOutputPalette (std::vector< te::rst::BandProperty::ColorEntry > &userPalette)
	Set the output user palette. More...

	~ClassifierKMeansStrategy ()

Protected Member Functions
bool	createOutputRaster (const std::vector< int > &bandsDataTypes, const std::vector< double > &noDataValues)
	Create the output raster using the EXPANSIBLE driver. More...

virtual void	reset ()
	Reset to an initial state. More...

void	setErrorMessage (const std::string &newErrorMessage)
	Set the current error message. More...

bool	setOutputRasterPalette (const unsigned int size)
	Create and set the output raster palette folowing the current internal settings. More...

Protected Attributes
bool	m_createRasterPalette
	Enable (true) or disable (false) the creation of a paletted output raster. More...

bool	m_enableMultiThread
	Enable or disable the use multipe threads. More...

bool	m_enableRasterCache
	Enable or disable the use a raster data cache. More...

std::vector< te::gm::Polygon * > const *	m_inputPolygonsPtr
	Input polygons. More...

std::vector< unsigned int >	m_inputRasterBands
	Input raster bands. More...

te::rst::Raster const *	m_inputRasterPtr
	A pointer to the input raster. More...

bool	m_isInitialized
	True if this instance is initialized. More...

std::map< std::string, std::string >	m_metaData
	Strategy-dependent metadata. More...

std::unique_ptr< te::rst::Raster >	m_outputRasterPtr
	A pointer to the output raster. More...

ClassifierKMeansStrategy::Parameters	m_parameters
	Internal execution parameters. More...

bool	m_progressInterfaceEnabled
	Progress interface status. More...

std::vector< te::rst::BandProperty::ColorEntry >	m_userOutputPalette
	User output raster palette (it must be large enough to accomodate all classifyier generated classes or an empty vector to automatically generate an random palette. More...

Detailed Description

KMeans strategy for image classification. Step-by-step:

Define a random solution, creating K values of mean.
Classify each pixel according to the closest mean.
Calculate new means using the classified pixels.
Go back to step 2, or stop depending on the following checks:
Check if the new means are different from the previous iteration, using the convergence threshold.
Check if maximum number of iterations has achieved.

Definition at line 52 of file ClassifierKMeansStrategy.h.

Constructor & Destructor Documentation

te::rp::ClassifierKMeansStrategy::ClassifierKMeansStrategy ( )

Definition at line 79 of file ClassifierKMeansStrategy.cpp.

References m_isInitialized, and ~ClassifierKMeansStrategy().

te::rp::ClassifierKMeansStrategy::~ClassifierKMeansStrategy ( )

default

Referenced by ClassifierKMeansStrategy().

Member Function Documentation

bool te::rp::ClassifierStrategy::createOutputRaster	(	const std::vector< int > &	bandsDataTypes,
		const std::vector< double > &	noDataValues
	)

protectedinherited

Create the output raster using the EXPANSIBLE driver.

Parameters

bandsDataTypes	Bands data types.
noDataValues	A vector of no-data values for each band.

Returns: true if ok,false on errors.

Note: A pointer to the created raster will be handled by m_outputRasterPtr.

Definition at line 105 of file ClassifierStrategy.cpp.

References te::rst::Raster::getBand(), te::rst::Raster::getGrid(), te::rst::Band::getProperty(), te::rst::GrayIdxCInt, te::sa::Grid, te::rst::BandProperty::m_colorInterp, te::rp::ClassifierStrategy::m_inputRasterBands, te::rp::ClassifierStrategy::m_inputRasterPtr, te::rst::BandProperty::m_noDataValue, te::rp::ClassifierStrategy::m_outputRasterPtr, te::rst::BandProperty::m_type, te::rst::RasterFactory::make(), and TERP_TRUE_OR_THROW.

Referenced by te::rp::ClassifierDummyStrategy::execute(), execute(), te::rp::ClassifierEDStrategy::execute(), te::rp::ClassifierMAPStrategy::execute(), te::rp::ClassifierEMStrategy::execute(), te::rp::ClassifierSAMStrategy::execute(), and te::rp::ClassifierISOSegStrategy::execute().

void te::rp::ClassifierStrategy::enableMultiThread ( const bool & enable )

inherited

Enable / disable the use of multiple threads.

Parameters

enable Enable (true) or disable (false) the use of multiple threads.

Definition at line 80 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_enableMultiThread.

void te::rp::ClassifierStrategy::enableOutputPalette ( const bool enabled )

inherited

Enable (true) or disable (false) the creation of a paletted output raster.

Parameters

enabled Enable (true) or disable (false) the creation of a paletted output raster.

Definition at line 90 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_createRasterPalette.

void te::rp::ClassifierStrategy::enableProgressInterface ( const bool & enable )

inherited

Enable / disable the progress interface.

Parameters

enable Enable (true) or disable (false) the progress interface.

Definition at line 70 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_progressInterfaceEnabled.

void te::rp::ClassifierStrategy::enableRasterCache ( const bool & enable )

inherited

Enable / disable the use of raster data cache.

Parameters

enable Enable (true) or disable (false) the use of raster cache.

Definition at line 75 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_enableRasterCache.

bool te::rp::ClassifierKMeansStrategy::execute	(		)
throw	(	te::rp::Exception
	)

virtual

Executes the classification strategy.

Returns: true if OK, false on errors.

Implements te::rp::ClassifierStrategy.

Definition at line 109 of file ClassifierKMeansStrategy.cpp.

References te::rst::RasterIterator< T >::begin(), te::rst::PointSetIterator< T >::begin(), te::cl::KMeans< TTRAIN, TCLASSIFY >::classify(), col, te::rp::ClassifierStrategy::createOutputRaster(), dt, te::rst::RasterIterator< T >::end(), te::rst::PointSetIterator< T >::end(), te::rst::Raster::getBand(), te::rst::RasterIterator< T >::getColumn(), te::rst::Raster::getNumberOfColumns(), te::rst::Raster::getNumberOfRows(), te::rst::Band::getProperty(), te::rst::GetRandomPointsInRaster(), te::rst::RasterIterator< T >::getRow(), te::rst::Band::getValue(), te::cl::KMeans< TTRAIN, TCLASSIFY >::initialize(), te::rp::ClassifierKMeansStrategy::Parameters::m_epsilon, te::rp::ClassifierStrategy::m_inputRasterBands, te::rp::ClassifierStrategy::m_inputRasterPtr, m_isInitialized, te::rp::ClassifierKMeansStrategy::Parameters::m_K, te::rp::ClassifierKMeansStrategy::Parameters::m_maxInputPoints, te::rp::ClassifierKMeansStrategy::Parameters::m_maxIterations, te::rst::BandProperty::m_noDataValue, te::rp::ClassifierStrategy::m_outputRasterPtr, m_parameters, te::common::TaskProgress::pulse(), te::common::TaskProgress::setCurrentStep(), te::common::TaskProgress::setMessage(), te::rp::ClassifierStrategy::setOutputRasterPalette(), te::common::TaskProgress::setTotalSteps(), te::rst::Band::setValue(), TE_TR, TERP_INSTANCE_TRUE_OR_RETURN_FALSE, te::cl::KMeans< TTRAIN, TCLASSIFY >::train(), and te::dt::UINT32_TYPE.

const std::string & te::rp::ClassifierStrategy::getErrorMessage ( ) const

inherited

Return the current error message if there is any.

Returns: Return the current error message if there is any.

Definition at line 100 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_errorMessage.

void te::rp::ClassifierStrategy::getMetaData ( std::map< std::string, std::string > & metadata ) const

inherited

Returns strategy-dependent metadata.

Parameters

metadata Strategy-dependent metadata.

Definition at line 85 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_metaData.

bool te::rp::ClassifierKMeansStrategy::initialize	(	ClassifierStrategyParameters const *const	strategyParams	)
throw	(	te::rp::Exception
	)

virtual

Initialize the classification strategy.

Parameters

strategyParams A pointer to the user given specific classification strategy parameters ou NULL if no parameters are present.

Returns: true if OK, false on errors.

Implements te::rp::ClassifierStrategy.

Definition at line 86 of file ClassifierKMeansStrategy.cpp.

References te::rp::ClassifierKMeansStrategy::Parameters::m_epsilon, m_isInitialized, te::rp::ClassifierKMeansStrategy::Parameters::m_K, te::rp::ClassifierKMeansStrategy::Parameters::m_maxInputPoints, te::rp::ClassifierKMeansStrategy::Parameters::m_maxIterations, m_parameters, TE_TR, and TERP_INSTANCE_TRUE_OR_RETURN_FALSE.

std::unique_ptr< te::rst::Raster > te::rp::ClassifierStrategy::releaseOutputRaster ( )

inherited

Returns a pointer to the output raster or a void pointer if there is none.

Returns: Returns a pointer to the output raster or a void pointer if there is none.

Note: The caller of this method must take the ownership of the returned object.

Definition at line 60 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_outputRasterPtr.

void te::rp::ClassifierStrategy::reset ( )

protectedvirtualinherited

Reset to an initial state.

Definition at line 185 of file ClassifierStrategy.cpp.

Referenced by te::rp::ClassifierStrategy::ClassifierStrategy(), and te::rp::ClassifierStrategy::operator=().

void te::rp::ClassifierStrategy::setErrorMessage ( const std::string & newErrorMessage )

protectedinherited

Set the current error message.

Parameters

newErrorMessage New error message;

Definition at line 180 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_errorMessage.

void te::rp::ClassifierStrategy::setInputPolygons ( const std::vector< te::gm::Polygon * > & polygonsPtrs )

inherited

Set the input polygons.

Parameters

polygonsPtrs Input polygons pointers.

Note: The caller of this method must keep the ownership of the objects.

Definition at line 65 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_inputPolygonsPtr.

void te::rp::ClassifierStrategy::setInputRaster ( const te::rst::Raster & raster )

inherited

Set the input raster.

Parameters

raster Input raster.

Definition at line 50 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_inputRasterPtr.

void te::rp::ClassifierStrategy::setInputRasterBands ( const std::vector< unsigned int > & rasterBands )

inherited

Set the input raste bandsr.

Parameters

rasterBands Input raster bands.

Definition at line 55 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_inputRasterBands.

bool te::rp::ClassifierStrategy::setOutputRasterPalette ( const unsigned int size )

protectedinherited

Create and set the output raster palette folowing the current internal settings.

Parameters

size	Palette size.

Returns: true if ok,false on errors.

Definition at line 145 of file ClassifierStrategy.cpp.

References te::rp::CreateFixedStepPalette(), te::rp::ClassifierStrategy::m_createRasterPalette, te::rp::ClassifierStrategy::m_outputRasterPtr, te::rp::ClassifierStrategy::m_userOutputPalette, te::rst::PaletteIdxCInt, and te::rst::RGBPalInt.

Referenced by te::rp::ClassifierDummyStrategy::execute(), execute(), te::rp::ClassifierEDStrategy::execute(), te::rp::ClassifierMAPStrategy::execute(), te::rp::ClassifierEMStrategy::execute(), te::rp::ClassifierSAMStrategy::execute(), and te::rp::ClassifierISOSegStrategy::execute().

void te::rp::ClassifierStrategy::setUserOutputPalette ( std::vector< te::rst::BandProperty::ColorEntry > & userPalette )

inherited

Set the output user palette.

Parameters

userPalette User output raster palette (it must be large enough to accomodate all classifyier generated classes or an empty vector to automatically generate an random palette.

Definition at line 95 of file ClassifierStrategy.cpp.

References te::rp::ClassifierStrategy::m_userOutputPalette.

Member Data Documentation

bool te::rp::ClassifierStrategy::m_createRasterPalette

protectedinherited

Enable (true) or disable (false) the creation of a paletted output raster.

Definition at line 164 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::enableOutputPalette(), te::rp::ClassifierStrategy::reset(), and te::rp::ClassifierStrategy::setOutputRasterPalette().

bool te::rp::ClassifierStrategy::m_enableMultiThread

protectedinherited

Enable or disable the use multipe threads.

Definition at line 159 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::enableMultiThread(), te::rp::ClassifierISOSegStrategy::execute(), and te::rp::ClassifierStrategy::reset().

bool te::rp::ClassifierStrategy::m_enableRasterCache

protectedinherited

Enable or disable the use a raster data cache.

Definition at line 154 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::enableRasterCache(), te::rp::ClassifierISOSegStrategy::execute(), and te::rp::ClassifierStrategy::reset().

std::vector<te::gm::Polygon*> const* te::rp::ClassifierStrategy::m_inputPolygonsPtr

protectedinherited

Input polygons.

Definition at line 179 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierDummyStrategy::execute(), te::rp::ClassifierISOSegStrategy::execute(), te::rp::ClassifierStrategy::reset(), and te::rp::ClassifierStrategy::setInputPolygons().

std::vector<unsigned int> te::rp::ClassifierStrategy::m_inputRasterBands

protectedinherited

Input raster bands.

Definition at line 174 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::createOutputRaster(), execute(), te::rp::ClassifierEDStrategy::execute(), te::rp::ClassifierEMStrategy::execute(), te::rp::ClassifierMAPStrategy::execute(), te::rp::ClassifierSAMStrategy::execute(), te::rp::ClassifierISOSegStrategy::execute(), te::rp::ClassifierStrategy::reset(), and te::rp::ClassifierStrategy::setInputRasterBands().

te::rst::Raster const* te::rp::ClassifierStrategy::m_inputRasterPtr

protectedinherited

A pointer to the input raster.

Definition at line 144 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::createOutputRaster(), execute(), te::rp::ClassifierEDStrategy::execute(), te::rp::ClassifierEMStrategy::execute(), te::rp::ClassifierMAPStrategy::execute(), te::rp::ClassifierSAMStrategy::execute(), te::rp::ClassifierISOSegStrategy::execute(), te::rp::ClassifierStrategy::reset(), and te::rp::ClassifierStrategy::setInputRaster().

bool te::rp::ClassifierKMeansStrategy::m_isInitialized

protected

True if this instance is initialized.

Definition at line 98 of file ClassifierKMeansStrategy.h.

Referenced by ClassifierKMeansStrategy(), execute(), and initialize().

std::map< std::string, std::string > te::rp::ClassifierStrategy::m_metaData

protectedinherited

Strategy-dependent metadata.

Definition at line 184 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::getMetaData(), and te::rp::ClassifierStrategy::reset().

std::unique_ptr< te::rst::Raster > te::rp::ClassifierStrategy::m_outputRasterPtr

protectedinherited

A pointer to the output raster.

Definition at line 169 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::createOutputRaster(), te::rp::ClassifierDummyStrategy::execute(), execute(), te::rp::ClassifierEDStrategy::execute(), te::rp::ClassifierEMStrategy::execute(), te::rp::ClassifierMAPStrategy::execute(), te::rp::ClassifierSAMStrategy::execute(), te::rp::ClassifierISOSegStrategy::execute(), te::rp::ClassifierStrategy::releaseOutputRaster(), te::rp::ClassifierStrategy::reset(), and te::rp::ClassifierStrategy::setOutputRasterPalette().

ClassifierKMeansStrategy::Parameters te::rp::ClassifierKMeansStrategy::m_parameters

protected

Internal execution parameters.

Definition at line 99 of file ClassifierKMeansStrategy.h.

Referenced by execute(), and initialize().

bool te::rp::ClassifierStrategy::m_progressInterfaceEnabled

protectedinherited

Progress interface status.

Definition at line 149 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::enableProgressInterface(), te::rp::ClassifierEDStrategy::execute(), te::rp::ClassifierMAPStrategy::execute(), te::rp::ClassifierSAMStrategy::execute(), te::rp::ClassifierISOSegStrategy::execute(), and te::rp::ClassifierStrategy::reset().

std::vector< te::rst::BandProperty::ColorEntry > te::rp::ClassifierStrategy::m_userOutputPalette

protectedinherited

User output raster palette (it must be large enough to accomodate all classifyier generated classes or an empty vector to automatically generate an random palette.

Definition at line 189 of file ClassifierStrategy.h.

Referenced by te::rp::ClassifierStrategy::reset(), te::rp::ClassifierStrategy::setOutputRasterPalette(), and te::rp::ClassifierStrategy::setUserOutputPalette().

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

Classes

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation