Collaboration diagram for Edge Builder:

Classes
class	vslam::EdgeBuilder< G >
	Interface for adding edges to the pose-graph. More...
class	vslam::MonoFtrClassEdgeBuilder< G >
	Implementation of EdgeBuilder, where there is only one kind of feature detector-descriptor pair. More...
Functions
static void	vslam::MonoFtrClassEdgeBuilder< G >::computeScale (VertexD contactvd, VertexD newvd, G &g, Points< G > &points, SE3 &scaled_T_vbd_vcd)
	Compute scale.
std::tr1::shared_ptr < RobustFeatureMatcher >	vslam::constructRMatcher (std::tr1::shared_ptr< cv::DescriptorMatcher > _dmatcher, std::tr1::shared_ptr< EMatrixModelBuilder > _minbuilder, std::tr1::shared_ptr< EMatrixModelBuilder > _builder)
static void	vslam::EdgeBuilder< G >::featureSelection (G &g, VertexD vd, int target_num, vector< int > &keep_idxs, vector< cv::KeyPoint > &sel_keypoints, cv::Mat &sel_descriptors)
	Feature Selection step for Edge Builder.
void	vslam::MonoFtrClassEdgeBuilder< G >::guidedMatching (G &g, VertexD c1vd, VertexD c2vd)
	Perform guided matching.
template<class ErrorAccumT >
ErrorAccumT::ErrorT	vslam::outlierRemoval (const Matrix3d &E, const ErrorAccumT &ea, const ImageMatchVec &matches_c1c2, const vector< cv::DMatch > &dmatches_c1c2, vector< cv::DMatch > &filtered_matches_c1c2)
	Outlier Removal for image correspondences in 2-view geometry given by an essential matrix.
template<class G >
void	vslam::setRelTranslationError (double r_min, double r_max, typename graph_traits< G >::vertex_descriptor vs, typename graph_traits< G >::vertex_descriptor vt, G &g)
bool	vslam::MonoFtrClassEdgeBuilder< G >::unguidedMatching (const vector< cv::KeyPoint > &sel_tvd_keypoints, const vector< cv::KeyPoint > &sel_qvd_keypoints, const cv::Mat &sel_tvd_descriptors, const cv::Mat &sel_qvd_descriptors, SE3 &T_tvd_qvd, std::vector< cv::DMatch > &global_dmatches_c1c2)
	Unguided Matching and Robust Model Estimations steps for Edge Builder.
Deprecated functions.
Deprecated: Old version of computing scale and finding new landmarks. Use MonoFtrClassEdgeBuilder::computeScale and findNewLandmarks instead.
static void	vslam::MonoFtrClassEdgeBuilder< G >::compute3viewStructure (VertexD vad, VertexD vbd, VertexD vcd, G &g, IdxPoint3dVec &, SE3 &)
	Compute scale and common points for two edges.
static void	vslam::MonoFtrClassEdgeBuilder< G >::matchScaleFindNewLandmarks (VertexD vsrc, VertexD vdest, Points< G > &points, G &g)
	Match scale for the new edge with the rest of the graph and find new landmarks.

Function Documentation

template<class G >

void vslam::MonoFtrClassEdgeBuilder< G >::compute3viewStructure	(	VertexD	vad,
		VertexD	vbd,
		VertexD	vcd,
		G &	g,
		IdxPoint3dVec &	common_pts_3d,
		SE3 &	scaled_T_vbd_vcd
	)		`[static, protected]`

Compute scale and common points for two edges.

For two edges sharing a vertex, compute feature track spanning both two edges. The tracks are represented as triangulated points and their indexes w.r.t. the c1 coordinate farme (i.e. vbd vertex).

Todo:: to be removed.

scaled_T_vbd_vcd SE3 transform with corrected scale as determined by common points with edge (vad, vbd).

Definition at line 280 of file graph_builder.h.

template<class G >

void vslam::MonoFtrClassEdgeBuilder< G >::computeScale	(	VertexD	convd,
		VertexD	newvd,
		G &	g,
		Points< G > &	points,
		SE3 &	scaled_T_vbd_vcd
	)		`[static, protected]`

Compute scale.

Todo:: document.

Parameters:

in/out] scaled_T_vbd_vcd SE3 transform with corrected scale as determined by feature tracks that are common with edge .

Todo:: force positive depth

Definition at line 409 of file graph_builder.h.

std::tr1::shared_ptr< RobustFeatureMatcher > vslam::constructRMatcher	(	std::tr1::shared_ptr< cv::DescriptorMatcher >	_dmatcher,
		std::tr1::shared_ptr< EMatrixModelBuilder >	_minbuilder,
		std::tr1::shared_ptr< EMatrixModelBuilder >	_builder
	)		`[inline]`

Definition at line 741 of file graph_builder.h.

template<class G >

void vslam::EdgeBuilder< G >::featureSelection	(	G &	g,
		VertexD	vd,
		int	target_num,
		vector< int > &	keep_idxs,
		vector< cv::KeyPoint > &	sel_keypoints,
		cv::Mat &	sel_descriptors
	)		`[static, protected]`

Feature Selection step for Edge Builder.

Selects approximately target_num features eavenly spread across the imaging surface.

Precondition:: vertex_kp and vertex_desc properties of the graph g are initialized correctly with detected keypoints. KeyPoint data structures in sel_keypoints have to have response parameter set.

Parameters:

[out]	keep_idxs	indexes of selected features in node vd.
[out]	sel_keypoints	array of cv::KeyPoint coresponding to the selected features in order given by keep_idxs.
[out]	sel_descriptors	same as sel_keypoints for cv::Mat feature descriptors.

Definition at line 517 of file graph_builder.h.

template<class G >

void vslam::MonoFtrClassEdgeBuilder< G >::guidedMatching	(	G &	g,
		VertexD	qvd,
		VertexD	tvd
	)		`[protected]`

Perform guided matching.

Performs guided matching by using vslam::findLocalMatches. The matches are then searched for outliers. Non-outlier matches that are not already in the feature-feature matches of edge '(qvd, tvd)', are then added.

Precondition:: see vslam::findLocalMatches

Todo:: unify errors under one variable as was done in rmatcher.

Definition at line 616 of file graph_builder.h.

template<class G >

void vslam::MonoFtrClassEdgeBuilder< G >::matchScaleFindNewLandmarks	(	VertexD	vsrc,
		VertexD	vdest,
		Points< G > &	points,
		G &	g
	)		`[static, protected]`

Match scale for the new edge with the rest of the graph and find new landmarks.

The function does two things. First, scale of the edge $(vsrc, vdest)$ is computed based on an edge connecting one of $vsrc$ and $vdest$ to the rest of the graph.

Second, new landmarks are created. New landmark is created if there exist two feature-feature matches sharing vsrc vertex, which are geometrically consistent and none of the vertices involved in the feature-feature matches is part of any existing landmark.

$vsrc$ and $vdest$ are not interchangable. If the edge connects unconnected vertex with the rest * of the graph, the unconnected vertex is passed as vdest and this method is called only once. Othervise it has to be called for both $(vsrc, vdest)$ and $(vdest, vsrc)$ edges.

Definition at line 342 of file graph_builder.h.

template<class ErrorAccumT >

ErrorAccumT::ErrorT vslam::outlierRemoval	(	const Matrix3d &	E,
		const ErrorAccumT &	ea,
		const ImageMatchVec &	matches_c1c2,
		const vector< cv::DMatch > &	dmatches_c1c2,
		vector< cv::DMatch > &	filtered_matches_c1c2
	)

Outlier Removal for image correspondences in 2-view geometry given by an essential matrix.

Parameters:

	E	essential matrix (c1 to c2).
	ea	ErrorAccumulator implementation (determines which data are outliers).
	matches_c1c2	image coordinates of matches from c1 to c2.
	dmatches_c1c2	feature index pairs corresponding to 'matches_c1c2'.
[out]	filtered_matches_c1c2	inlier feature index pairs from 'dmatches_c1c2'

Returns:: model error as determined by 'ea'.

Definition at line 585 of file graph_builder.h.

template<class G >

void vslam::setRelTranslationError	(	double	r_min,
		double	r_max,
		typename graph_traits< G >::vertex_descriptor	vs,
		typename graph_traits< G >::vertex_descriptor	vt,
		G &	g
	)

Definition at line 249 of file graph_builder.h.

template<class G >

bool vslam::MonoFtrClassEdgeBuilder< G >::unguidedMatching	(	const vector< cv::KeyPoint > &	sel_tvd_keypoints,
		const vector< cv::KeyPoint > &	sel_qvd_keypoints,
		const cv::Mat &	sel_tvd_descriptors,
		const cv::Mat &	sel_qvd_descriptors,
		SE3 &	T_tvd_qvd,
		std::vector< cv::DMatch > &	global_dmatches_c1c2
	)		`[protected]`

Unguided Matching and Robust Model Estimations steps for Edge Builder.

Parameters:

	sel_tvd_keypoints	target vertex keypoints array.
	sel_qvd_keypoints	query vertex keypoints array.
	sel_tvd_descriptors	target vertex feature descriptors array.
	sel_qvd_descriptors	query vertex feature descriptors array.
[out]	T_tvd_qvd	RBT from target vertex CF to query vertex CF with undefined scale.
[out]	global_dmatches_c1c2	computed matches.

Returns:: true if T_tvd_qvd contains valid value and false othervise.

Definition at line 547 of file graph_builder.h.

Classes

Functions

Deprecated functions.

Function Documentation