blortRecognizer::Recognizer3D Class Reference

Recognizer for 3D pose estimation using SIFT features. More...

#include <Recognizer3D.h>

Public Member Functions
unsigned int	getCodeBookSize ()
cv::Mat	getDebugImage ()
cv::Mat	getImage ()
void	getLastSifts (std::vector< Siftex > &sl)
	get position and normal vector of sift features in model detected in the last call of learnSifts()
void	getSifts (std::vector< Siftex > &sl)
	get position and normal vector of sift features in model
bool	learnSifts (IplImage *tFrame, const TomGine::tgModel &model, const TomGine::tgPose &pose)
	add sift features to sift model of an object
bool	loadModelFromFile (const std::string sift_file)
	load a sift model
bool	recognize (IplImage *tFrame, TomGine::tgPose &pose, float &conf)
	recognizes a object by using the loaded sift model file
	Recognizer3D (const CameraParameter &camParam, std::string config_root="", bool display=false)
	Construction of Recognizer.
bool	saveModelToFile (const char *sift_file)
	save a sift model
void	setCameraParameter (const blortRecognizer::CameraParameter &camParam)
void	setDoUndistort (bool enable)
void	setGaussian (int kernel, float stdDeviation)
void	setNNThreshold (double nn_threshold)
void	setRansacNPointsToMatch (unsigned int n)
	~Recognizer3D ()
Static Public Member Functions
static void	Convert (P::PoseCv &p1, TomGine::tgPose &p2)
Private Member Functions
	Recognizer3D ()
Private Attributes
CvMat *	C
std::string	config_root
cv::Mat	debug_image
cv::Mat	display_image
bool	do_undistort
CameraParameter	m_cp
P::ODetect3D	m_detect
bool	m_display
float	m_gauss_dev
int	m_gauss_kernel
P::Array< P::KeypointDescriptor * >	m_image_keys
std::vector< Siftex >	m_lastsiftexlist
bool	m_model_loaded
P::Object3D	m_sift_model
P::ModelObject3D	m_sift_model_learner
std::vector< Siftex >	m_siftexlist
CvMat *	pDistortion
CvMat *	pIntrinsicDistort
CvMat *	pMapX
CvMat *	pMapY
P::DetectGPUSIFT	sift

Detailed Description

Recognizer for 3D pose estimation using SIFT features.

Definition at line 47 of file Recognizer3D.h.

Constructor & Destructor Documentation

Recognizer3D::Recognizer3D	(	const CameraParameter &	camParam,
		std::string	config_root = `""`,
		bool	display = `false`
	)

Construction of Recognizer.

Parameters:

	camParam	camera parameter for camera calibration
	prefix	to be added to every file path
	display	display sifts and found object in a window

Definition at line 47 of file Recognizer3D.cpp.

Recognizer3D::~Recognizer3D ( )

Definition at line 109 of file Recognizer3D.cpp.

blortRecognizer::Recognizer3D::Recognizer3D ( ) [private]

Member Function Documentation

void Recognizer3D::Convert	(	P::PoseCv &	p1,
		TomGine::tgPose &	p2
	)			`[static]`

Definition at line 31 of file Recognizer3D.cpp.

unsigned int Recognizer3D::getCodeBookSize ( )

Definition at line 377 of file Recognizer3D.cpp.

cv::Mat blortRecognizer::Recognizer3D::getDebugImage ( ) [inline]

Definition at line 104 of file Recognizer3D.h.

cv::Mat blortRecognizer::Recognizer3D::getImage ( ) [inline]

Definition at line 97 of file Recognizer3D.h.

void blortRecognizer::Recognizer3D::getLastSifts ( std::vector< Siftex > & sl ) [inline]

get position and normal vector of sift features in model detected in the last call of learnSifts()

Parameters:

	pl	3D point list of sift features (relative to object)
	nl	3D normal vector list of sift features (in object space)

Definition at line 93 of file Recognizer3D.h.

void blortRecognizer::Recognizer3D::getSifts ( std::vector< Siftex > & sl ) [inline]

get position and normal vector of sift features in model

Parameters:

	pl	3D point list of sift features (relative to object)
	nl	3D normal vector list of sift features (in object space)

Definition at line 88 of file Recognizer3D.h.

bool Recognizer3D::learnSifts	(	IplImage *	tFrame,
		const TomGine::tgModel &	model,
		const TomGine::tgPose &	pose
	)

add sift features to sift model of an object

Parameters:

	tFrame	image/pixel map to search for new sift features
	model	shape description of model using faces and vertices (see TomGine::tgModel)
	pose	pose of the model as seen in the image