#include <recognitionmodel.h>

Public Member Functions
int	calculate (const cv::Mat &image, const PointCloud::Ptr &pointcloud)
void	findCorrespondences (const ObjectAspect &other, std::vector< cv::Point2i > &ptpairs)
void	fromFile (const std::string &filename)
double	matchAspects (ObjectAspect &other, Eigen::Matrix4f &transform)
	ObjectAspect ()
void	plotKeypoints (cv::Mat &image)
Static Public Member Functions
static double	compareSURFDescriptors (const float d1, const float d2, double best, int length)
static int	naiveNearestNeighbor (const float vec, const std::vector< cv::KeyPoint > &model_keypoints, const float model_descriptors, const std::map< int, int > &imageMap2D3D)
Public Attributes
std::vector< float >	descriptors
	SURF descriptors.
cv::Mat	image
	2D input image
std::vector< cv::KeyPoint >	keypoints
	2D keypoints
PointCloud::Ptr	keypoints3D
	3D keypoints
std::map< int, int >	map2D3D
	maps 2D keypoints to 3D coordinates (indices from keypoints to indices from keypoints3D)
std::map< int, int >	map2D3Dinv
	inverse to map2D3D
ObjectAspect *	match
PointCloud::Ptr	points
	input pointcloud

Detailed Description

An object aspect corresponds to a single view of the model. It stores a set of SURF keypoints and descriptors along with 3D information for these. Recognition models consist of multiple aspects. When detecting an object, we go through all object aspects and try to find the best match.

Definition at line 57 of file recognitionmodel.h.

Constructor & Destructor Documentation

ObjectAspect::ObjectAspect ( )

Definition at line 59 of file recognitionmodel.cpp.

Member Function Documentation

int ObjectAspect::calculate	(	const cv::Mat &	image,
		const PointCloud::Ptr &	pointcloud
	)

Calculates the SURF keypoints and descriptors for this aspect and determines their 3D coordinates.

Parameters:

image	2D camera image
pointcloud	dense point cloud for extracting 3D coordinates

Definition at line 64 of file recognitionmodel.cpp.

double ObjectAspect::compareSURFDescriptors	(	const float *	d1,
		const float *	d2,
		double	best,
		int	length
	)		`[static]`

Calculates the distance between 2 SURF descriptors.

Returns:: euclidean distance

Parameters:

d1	first descriptor
d2	second descriptor
best	lowest distance found until now
length	length of descriptors (has to be a multiple of 4)

Definition at line 277 of file recognitionmodel.cpp.

void ObjectAspect::findCorrespondences	(	const ObjectAspect &	other,
		std::vector< cv::Point2i > &	ptpairs
	)

Finds a set of correspondences between this and another object aspect and gives a list of correspondences.

Parameters:

other	the other ObjectAspect to compare
ptpairs	output parameter, where each cv::Point2i contains (index of this ObjectAspect's keypoint, index of other's keypoint)

Definition at line 293 of file recognitionmodel.cpp.

void ObjectAspect::fromFile ( const std::string & filename )

Loads an ObjectAspect from a dense .pcd file.

Parameters:

filename filename to a dense .pcd

Definition at line 407 of file recognitionmodel.cpp.

double ObjectAspect::matchAspects	(	ObjectAspect &	other,
		Eigen::Matrix4f &	transform
	)

Tries to match 2 aspects and calculates the resulting transformation.

Returns:: 1 for success, DBL_MAX for failure

Parameters:

other	the ObjectAspect to compare with
output	parameter, gives the resulting transformation

Definition at line 115 of file recognitionmodel.cpp.

int ObjectAspect::naiveNearestNeighbor	(	const float *	vec,
		const std::vector< cv::KeyPoint > &	model_keypoints,
		const float *	model_descriptors,
		const std::map< int, int > &	imageMap2D3D
	)		`[static]`

Finds the nearest neighbor of a given SURF descriptor.

Returns:: index of the nearest neighbor, or -1 if none is found

Parameters:

vec	SURF descriptor
model_keypoints	keypoints of the ObjectAspect
model_descriptors	descriptors of the ObjectAspect
imageMap2D3D	map from 2D points to 3D points