jafar::bundle::bundleLinearSys Class Reference

Mathematical part of our Bundle Adjustment implementation. More...

---

Detailed Description

Mathematical part of our Bundle Adjustment implementation.

Definition at line 27 of file linearsys.hpp.

#include <linearsys.hpp>

List of all members.

Public Member Functions
	bundleLinearSys (const observationsContainer obsCon_)
	Default constructor.
void	afficheB1 (unsigned int indObs, unsigned int indVue)
void	afficheB2 (unsigned int indObs, unsigned int indObj)
template<typename typeCam >
void	processData (bundle::observationsContainer &bobs, bundle::bundleVariables< typeCam > &bvar)
	Computation of the jacobian matrix J, of the gradient vector G and of the matrix A=trans(J)*SIGMA*J (see Lourakis's TR to more infos).
double	maxDiagA ()
	Return the max value of the diagonal of A.
void	solveByLU (double mu=0.0)
	Resolve the linear system (A+muId)dp = g by LU decomposition.
void	solveByGC (double mu=0.0)
	Resolve the linear system (A+muId)dp = g by the Conjugate Gradient method.
Public Attributes
jblas::vec	gradient
	Gradient vector.
jblas::vec	deltap
	Incremental upate of the BA variables vector (extrinsic parameters + 3D points localization)
Private Member Functions
void	bundle_forward_substitution (double const mu)
void	bundle_backward_substitution ()
	After resvoling of S dpCPDV = T, dp3D is obtained by backward substitution.
Private Attributes
unsigned int	nbCam
unsigned int	nbObj3D
unsigned int	nbObs2D
jmath::matofmat< double, 2, 6 >	B1
jmath::matofmat< double, 2, 3 >	B2
std::vector< jblas::mat66 >	U
std::vector< jblas::mat33 >	V
std::vector< jblas::mat33 >	invV
jblas::mat	Y
std::vector< jblas::mat_range * >	rangeY
jblas::vec	T
std::vector< jblas::vec_range * >	rangeT
jblas::mat	S
std::vector< jblas::mat_range * >	rangeS
jblas::mat	D
std::vector< jblas::mat_range * >	rangeD
std::vector< jblas::vec_range * >	gradCpdv
std::vector< jblas::vec_range * >	grad3d
jblas::vec_range	dpcpdv
jblas::vec_range	dp3d
std::vector< jblas::vec_range * >	dp1cpdv
std::vector< jblas::vec_range * >	dp1obj3d

---

Member Function Documentation

double jafar::bundle::bundleLinearSys::maxDiagA

(

)

[inline]

Return the max value of the diagonal of A.

max(Aii) = max(max(Uii),max(V(ii))) This value is used to initialize mu (in the Levenberg Marquardt algorithm)

Definition at line 344 of file linearsys.hpp.

Referenced by jafar::bundle::bundleLM().

---

The documentation for this class was generated from the following file:

• modules/bundle/include/bundle/linearsys.hpp