Package boofcv.alg.structure

Class PairwiseGraphUtils

java.lang.Object

boofcv.alg.structure.PairwiseGraphUtils

public class PairwiseGraphUtils
extends Object

Various utility functions for dealing with PairwiseImageGraph

Field Summary

Fields

Modifier and Type	Field	Description
final DogArray_I32	commonIdx
ConfigConverge	configConvergeSBA	Converge tolerance for SBA
boolean	configScaleSBA	Toggles scaling inputs for SBA
LookUpCameraInfo	dbCams
LookUpSimilarImages	dbSimilar
final DogArray<Point2D_F64>	featsA
final DogArray<Point2D_F64>	featsB
final DogArray<Point2D_F64>	featsC
final DogArray_I32	inlierIdx	Which features the inliers correspond to
final FastArray<AssociatedTriple>	inliersThreeView	Inliers from robust fitting of trifocal tensor.
final DogArray<AssociatedTriple>	matchesTriple
protected final SceneObservations	observations
final DMatrixRMaj	P1	Projective camera matrices for 3-View reconstruction.
final DMatrixRMaj	P2
final DMatrixRMaj	P3
protected PoseFromPairLinear6	poseEstimator
CameraPinholeBrown	priorCamA	Prior calibration information for each view's camera
CameraPinholeBrown	priorCamB
CameraPinholeBrown	priorCamC
protected ModelMatcher<TrifocalTensor,AssociatedTriple>	ransac
protected BundleAdjustment<SceneStructureProjective>	sba
protected ScaleSceneStructure	scaler
PairwiseImageGraph.View	seed	The three views used in three view algorithms
protected final SceneStructureProjective	structurePr	The estimated scene structure.
final DogArray_I32	table_A_to_B
final DogArray_I32	table_A_to_C
final DogArray_I32	table_B_to_C
final DogArray_I32	table_C_to_A
protected TriangulateNViewsProjective	triangulator
PairwiseImageGraph.View	viewB	The three views used in three view algorithms
PairwiseImageGraph.View	viewC	The three views used in three view algorithms
DogArray_B	visibleAll
DogArray_B	visibleMotion

Constructor Summary

Constructors

Constructor	Description
PairwiseGraphUtils()
PairwiseGraphUtils(TriangulateNViewsProjective triangulator, ModelMatcher<TrifocalTensor,AssociatedTriple> ransac, BundleAdjustment<SceneStructureProjective> sba)
PairwiseGraphUtils(ConfigProjectiveReconstruction config)

Method Summary

Modifier and Type	Method	Description
static void	createTableViewAtoB(PairwiseImageGraph.View viewA, PairwiseImageGraph.Motion edge, DogArray_I32 table_a_to_b)	Creates a look up table for converting features indexes from view A to view B from the list of inliers.
void	createThreeViewLookUpTables()	Creates three feature look up tables: A -> B, A -> C, and B -> C
void	createTripleFromCommon(@Nullable PrintStream verbose)	Convert triple from indexes into coordinates
boolean	estimateProjectiveCamerasRobustly()	Robustly estimates the trifocal tensor and extracts canonical camera matrices
void	findAllConnectedSeed(PairwiseImageGraph.View seed, DogArray_I32 connectIdx, DogArray_I32 commonIdx)	Finds the indexes of tracks which are common to all views and are inliers.
void	findFullyConnectedTriple()	Finds which features are common between all three views using the look up tables.
void	findFullyConnectedTriple(DogArray_I32 selectedIdxA)	Same as findFullyConnectedTriple() but with a list that specified a subset of features in seed.
void	initializeSbaObservationsThreeView()	Converts the set of inlier observations from the three view inliers into observations that bundle adjustment can understand
void	initializeSbaSceneThreeView(boolean fixedSeed)	Initializes projective reconstruction from 3-views.
boolean	refineWithBundleAdjustment()	Last step is to refine the current initial estimate with bundle adjustment
SceneWorkingGraph.InlierInfo	saveRansacInliers(SceneWorkingGraph.View view, DogArray_I32 inlierIdx)	Saves which features were used as inliers.
protected void	triangulateFeatures()	Triangulates the location of each features in homogenous space and save to bundle adjustment scene

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

dbSimilar

public LookUpSimilarImages dbSimilar

dbCams

public LookUpCameraInfo dbCams

visibleAll

public DogArray_B visibleAll

visibleMotion

public DogArray_B visibleMotion

configConvergeSBA

public ConfigConverge configConvergeSBA

Converge tolerance for SBA

configScaleSBA

public boolean configScaleSBA

Toggles scaling inputs for SBA

structurePr

protected final SceneStructureProjective structurePr

The estimated scene structure. This the final estimated scene state

observations

protected final SceneObservations observations

ransac

protected ModelMatcher<TrifocalTensor,AssociatedTriple> ransac

triangulator

protected TriangulateNViewsProjective triangulator

poseEstimator

protected PoseFromPairLinear6 poseEstimator

sba

protected BundleAdjustment<SceneStructureProjective> sba

scaler

protected ScaleSceneStructure scaler

seed

public PairwiseImageGraph.View seed

The three views used in three view algorithms

viewB

public PairwiseImageGraph.View viewB

The three views used in three view algorithms

viewC

public PairwiseImageGraph.View viewC

The three views used in three view algorithms

priorCamA

public CameraPinholeBrown priorCamA

Prior calibration information for each view's camera

priorCamB

public CameraPinholeBrown priorCamB

priorCamC

public CameraPinholeBrown priorCamC

matchesTriple

public final DogArray<AssociatedTriple> matchesTriple

inliersThreeView

public final FastArray<AssociatedTriple> inliersThreeView

Inliers from robust fitting of trifocal tensor. Pixels

inlierIdx

public final DogArray_I32 inlierIdx

Which features the inliers correspond to

P1

public final DMatrixRMaj P1

Projective camera matrices for 3-View reconstruction. P1 is always identity

P2

public final DMatrixRMaj P2

P3

public final DMatrixRMaj P3

featsA

public final DogArray<Point2D_F64> featsA

featsB

public final DogArray<Point2D_F64> featsB

featsC

public final DogArray<Point2D_F64> featsC

table_A_to_B

public final DogArray_I32 table_A_to_B

table_A_to_C

public final DogArray_I32 table_A_to_C

table_B_to_C

public final DogArray_I32 table_B_to_C

table_C_to_A

public final DogArray_I32 table_C_to_A

commonIdx

public final DogArray_I32 commonIdx

Constructor Details

PairwiseGraphUtils

public PairwiseGraphUtils(TriangulateNViewsProjective triangulator,
 ModelMatcher<TrifocalTensor,AssociatedTriple> ransac,
 BundleAdjustment<SceneStructureProjective> sba)

PairwiseGraphUtils

public PairwiseGraphUtils(ConfigProjectiveReconstruction config)

PairwiseGraphUtils

public PairwiseGraphUtils()

Method Details

findAllConnectedSeed

public void findAllConnectedSeed(PairwiseImageGraph.View seed,
 DogArray_I32 connectIdx,
 DogArray_I32 commonIdx)

Finds the indexes of tracks which are common to all views and are inliers.

Parameters:

seed - (Input) The view which is used as the reference point

connectIdx - (Input) Indexes of connections in the seed view that will be searched for common connections

commonIdx - (Output) Indexes of observation in seed that are visible in connected views

createThreeViewLookUpTables

public void createThreeViewLookUpTables()

Creates three feature look up tables: A -> B, A -> C, and B -> C

findFullyConnectedTriple

public void findFullyConnectedTriple()

Finds which features are common between all three views using the look up tables. Results are stored in commonIdx by index of feature in seed.

findFullyConnectedTriple

public void findFullyConnectedTriple(DogArray_I32 selectedIdxA)

Same as findFullyConnectedTriple() but with a list that specified a subset of features in seed.

Parameters:

selectedIdxA - List of feature indexes in A to consider

createTripleFromCommon

public void createTripleFromCommon(@Nullable
 @Nullable PrintStream verbose)

Convert triple from indexes into coordinates

estimateProjectiveCamerasRobustly

public boolean estimateProjectiveCamerasRobustly()

Robustly estimates the trifocal tensor and extracts canonical camera matrices

saveRansacInliers

public SceneWorkingGraph.InlierInfo saveRansacInliers(SceneWorkingGraph.View view,
 DogArray_I32 inlierIdx)

Saves which features were used as inliers.

Parameters:

inlierIdx - Which features in 'commonIdx' are inliers and should be added to the set

view - Which view should be the first view in the list and have its inliers updated.

initializeSbaSceneThreeView

public void initializeSbaSceneThreeView(boolean fixedSeed)

Initializes projective reconstruction from 3-views. 1) RANSAC to fit a trifocal tensor 2) Extract camera matrices that have a common projective space 3) Triangulate location of 3D homogenous points

Parameters:

fixedSeed - if true the seed is fixed and other views are not. If false then the inverse happens.

triangulateFeatures

protected void triangulateFeatures()

Triangulates the location of each features in homogenous space and save to bundle adjustment scene

initializeSbaObservationsThreeView

public void initializeSbaObservationsThreeView()

Converts the set of inlier observations from the three view inliers into observations that bundle adjustment can understand

refineWithBundleAdjustment

public boolean refineWithBundleAdjustment()

Last step is to refine the current initial estimate with bundle adjustment

createTableViewAtoB

public static void createTableViewAtoB(PairwiseImageGraph.View viewA,
 PairwiseImageGraph.Motion edge,
 DogArray_I32 table_a_to_b)

Creates a look up table for converting features indexes from view A to view B from the list of inliers.

Parameters:

viewA - (input) The view that is the src in the table

edge - (input) Edge connect view A to B, this contains the dst in the table

table_a_to_b - (output) The resulting lookup table