ipa_PeopleDetector::PeopleDetector Class Reference

#include <PeopleDetector.h>

Public Member Functions
virtual unsigned long	AddFace (cv::Mat &img, cv::Rect &face, std::string id, std::vector< cv::Mat > &images, std::vector< std::string > &ids)
virtual unsigned long	CalculateFaceClasses (cv::Mat &projectedTrainFaceMat, std::vector< std::string > &id, int nEigens, cv::Mat &faceClassAvgProjections, std::vector< std::string > &idUnique, cv::SVM personClassifier=0)
virtual unsigned long	ClassifyFace (float projectedTestFace, int nearest, int nEigens, cv::Mat &projectedTrainFaceMat, int threshold, cv::Mat &eigenValMat, cv::SVM *personClassifier=0)
virtual unsigned long	ConvertAndResize (cv::Mat &img, cv::Mat &resized, cv::Rect &face)
virtual unsigned long	DetectColorFaces (cv::Mat &img, std::vector< cv::Rect > &faceCoordinates)
virtual unsigned long	DetectFaces (cv::Mat &img, cv::Mat &rangeImg, std::vector< cv::Rect > &colorFaceCoordinates, std::vector< cv::Rect > &rangeFaceCoordinates, std::set< size_t > &colorToRangeFaceDependency, bool fillUnassignedDepthValues=false)
virtual unsigned long	DetectRangeFace (cv::Mat &img, std::vector< cv::Rect > &rangeFaceCoordinates, bool fillUnassignedDepthValues=false)
virtual unsigned long	Init (std::string directory)
virtual unsigned long	PCA (int *nEigens, std::vector< cv::Mat > &eigenVectors, cv::Mat &eigenValMat, cv::Mat &avgImage, std::vector< cv::Mat > &images, cv::Mat &projectedTrainFaceMat)
	PeopleDetector (void)
	Constructor.
virtual cv::Mat	preprocessImage (cv::Mat &input_image)
virtual unsigned long	RecognizeFace (cv::Mat &colorImage, std::vector< cv::Rect > &colorFaces, int nEigens, std::vector< cv::Mat > &eigenVectArr, cv::Mat &avgImage, cv::Mat &projectedTrainFaceMat, std::vector< int > &index, int threshold, int threshold_FS, cv::Mat &eigenValMat, cv::SVM personClassifier=0)
	~PeopleDetector (void)
	Destructor.
Public Attributes
int	m_faces_drop_groups
	Minimum number (minus 1) of neighbor rectangles that makes up an object.
double	m_faces_increase_search_scale
	The factor by which the search window is scaled between the subsequent scans.
int	m_faces_min_search_scale_x
	Minimum search scale x.
int	m_faces_min_search_scale_y
	Minimum search scale y.
int	m_range_drop_groups
	Minimum number (minus 1) of neighbor rectangles that makes up an object.
double	m_range_increase_search_scale
	The factor by which the search window is scaled between the subsequent scans.
int	m_range_min_search_scale_x
	Minimum search scale x.
int	m_range_min_search_scale_y
	Minimum search scale y.
Private Member Functions
unsigned long	InterpolateUnassignedPixels (cv::Mat &img)
Private Attributes
CvHaarClassifierCascade *	m_face_cascade
	Haar-Classifier for face-detection.
CvHaarClassifierCascade *	m_range_cascade
	Haar-Classifier for range-detection.
CvMemStorage *	m_storage
	Storage for face and eye detection.

Detailed Description

Interface to Calibrate Head of Care-O-bot 3. Long description

Definition at line 25 of file PeopleDetector.h.

Constructor & Destructor Documentation

PeopleDetector::PeopleDetector ( void )

Constructor.

Constructor

Definition at line 16 of file PeopleDetector.cpp.

PeopleDetector::~PeopleDetector ( void )

Destructor.

Definition at line 38 of file PeopleDetector.cpp.

Member Function Documentation

unsigned long PeopleDetector::AddFace	(	cv::Mat &	img,
		cv::Rect &	face,
		std::string	id,
		std::vector< cv::Mat > &	images,
		std::vector< std::string > &	ids
	)			`[virtual]`

Function to add a new face The function adds a new face to the trained images

Parameters:

	img	Image
	face	The face
	id	Id of the new face
	images	Vector with trained images
	ids	Vector with trained ids

Returns:: Return code

Definition at line 201 of file PeopleDetector.cpp.

unsigned long PeopleDetector::CalculateFaceClasses	(	cv::Mat &	projectedTrainFaceMat,
		std::vector< std::string > &	id,
		int *	nEigens,
		cv::Mat &	faceClassAvgProjections,
		std::vector< std::string > &	idUnique,
		cv::SVM *	personClassifier = `0`
	)			`[virtual]`

Function to calculate the FaceClasses The function calculates the average eigenvector decomposition factors for each face classes.

Parameters:

	projectedTrainFaceMat	The projected training faces
	id	The ids of the training faces
	nEigens	Number of eigenvalues
	faceClassAvgProjections	The average factors of the eigenvector decomposition from each face class
	idUnique	A vector containing all different Ids from the training session exactly once (idUnique[i] stores the corresponding id to the average face coordinates in the face subspace in faceClassAvgProjections.row(i))
	personClassifier	A classifier for person identification. It is trained in this function. Can be left out if a simpler identification method is used.

Returns:: Return code

Definition at line 466 of file PeopleDetector.cpp.

unsigned long PeopleDetector::ClassifyFace	(	float *	projectedTestFace,
		int *	nearest,
		int *	nEigens,
		cv::Mat &	projectedTrainFaceMat,
		int *	threshold,
		cv::Mat &	eigenValMat,
		cv::SVM *	personClassifier = `0`
	)			`[virtual]`

Function to find the closest face class The function calculates the distance of each sample image to the trained face class

Parameters:

	projectedTestFace	The projected test face
	nearest	Index of nearest face, or -1 if the face is unknown
	nEigens	Number of eigenvalues
	projectedTrainFaceMat	The average factors from each face class originating from the eigenvector decomposition
	threshold	The threshold to recognize unkown faces
	eigenValMat	Eigenvalues
	personClassifier	A classifier for person identification. It is trained in this function. Can be left out if a simpler identification method is used.

Returns:: Return code

Definition at line 404 of file PeopleDetector.cpp.

unsigned long PeopleDetector::ConvertAndResize	(	cv::Mat &	img,
		cv::Mat &	resized,
		cv::Rect &	face
	)			`[virtual]`

Function to Convert and Resizes a given image The function converts a 8U3 image from camera to an 8U1 image and resizes the face to 100x100 px.

Parameters:

	img	Image from camera
	resized	Resized image from face
	face	The face

Returns:: Return code

Definition at line 214 of file PeopleDetector.cpp.

unsigned long PeopleDetector::DetectColorFaces	(	cv::Mat &	img,
		std::vector< cv::Rect > &	faceCoordinates
	)			`[virtual]`

Function to detect the faces on color image The function detects the faces in an given image

Parameters:

	img	Image
	faceCoordinates	Vector with the coordinates of detected faces in color image

Returns:: Return code

Definition at line 46 of file PeopleDetector.cpp.

unsigned long PeopleDetector::DetectFaces	(	cv::Mat &	img,
		cv::Mat &	rangeImg,
		std::vector< cv::Rect > &	colorFaceCoordinates,
		std::vector< cv::Rect > &	rangeFaceCoordinates,
		std::set< size_t > &	colorToRangeFaceDependency,
		bool	fillUnassignedDepthValues = `false`
	)			`[virtual]`

Function to detect faces The function calls internally the functions DetectRangeFace() and DetectColorFaces()

Parameters:

	img	Color image
	rangeImg	Range image
	faceCoordinates	Vector with the coordinates of detected faces on complete color image
	rangeFaceCoordinates	Vector with the coordinates of heads on range image
	colorToRangeFaceDependency	stores the indices of range images that contain a face detection in the color image
	fillUnassignedDepthValues	this parameter should be true if the kinect sensor is used (activates a filling method for black pixels)

Returns:: Return code

Definition at line 136 of file PeopleDetector.cpp.

unsigned long PeopleDetector::DetectRangeFace	(	cv::Mat &	img,
		std::vector< cv::Rect > &	rangeFaceCoordinates,
		bool	fillUnassignedDepthValues = `false`
	)			`[virtual]`

Function to detect the face on range image The function detects the face in an given range image

Parameters:

	img	Image
	rangeFaceCoordinates	Vector with the coordinates of detected heads in range image
	fillUnassignedDepthValues	this parameter should be true if the kinect sensor is used (activates a filling method for black pixels)

Returns:: Return code

Definition at line 107 of file PeopleDetector.cpp.

unsigned long PeopleDetector::Init ( std::string directory ) [virtual]

Initialization function. Creates an instance of a range imaging sensor (i.e. SwissRanger SR-3000) and an instance of

Parameters:

directory

The directory for data files

Returns:: Return code

Definition at line 22 of file PeopleDetector.cpp.

unsigned long PeopleDetector::InterpolateUnassignedPixels ( cv::Mat & img ) [private]

interpolates unassigned pixels in the depth image when using the kinect

Parameters:

img

depth image

Returns:: Return code

Definition at line 73 of file PeopleDetector.cpp.

unsigned long PeopleDetector::PCA	(	int *	nEigens,
		std::vector< cv::Mat > &	eigenVectors,
		cv::Mat &	eigenValMat,
		cv::Mat &	avgImage,
		std::vector< cv::Mat > &	images,
		cv::Mat &	projectedTrainFaceMat
	)			`[virtual]`

Function to Run the PCA algorithm

Parameters:

	nEigens	Number of eigenvalues
	eigenVectors	Eigenvectors
	eigenValMat	Eigenvalues
	avgImage	Average image
	images	Trained faces
	projectedTrainFaceMat	Projected training faces (coefficients for the eigenvectors of the face subspace)

Returns:: Return code

Definition at line 267 of file PeopleDetector.cpp.

cv::Mat PeopleDetector::preprocessImage ( cv::Mat & input_image ) [virtual]

Applies some preprocessing to the grayscale face images to obtain a more robust identification.

Parameters:

input_image

Grayscale face image.

Returns:: Preprocessed image.

Definition at line 224 of file PeopleDetector.cpp.

unsigned long PeopleDetector::RecognizeFace	(	cv::Mat &	colorImage,
		std::vector< cv::Rect > &	colorFaces,
		int *	nEigens,
		std::vector< cv::Mat > &	eigenVectArr,
		cv::Mat &	avgImage,
		cv::Mat &	projectedTrainFaceMat,
		std::vector< int > &	index,
		int *	threshold,
		int *	threshold_FS,
		cv::Mat &	eigenValMat,
		cv::SVM *	personClassifier = `0`
	)			`[virtual]`

Function to Recognize faces The function recognize the faces

Parameters:

	colorImage	source color image
	faceCoordinates	Detected faces
	nEigens	Number of eigenvalues
	eigenVectArr	Eigenvectors
	avgImage	Average image
	projectedTrainFaceMat	Projected training faces
	index	Index of classified face in vector
	threshold	The threshold to recognize unkown faces
	threshold_FS	The threshold to the face space
	eigenValMat	Eigenvalues
	personClassifier	A classifier for person identification. It is trained in this function. Can be left out if a simpler identification method is used.