gtsam
Loading...
Searching...
No Matches
gtsam::SelfCalibrationFactor Class Reference

Detailed Description

Binary factor measuring how close a fundamental matrix is to admitting a valid essential matrix, as a function of the two cameras' focal lengths.

Implements the focal-length self-calibration residual of Fetzer et al. (WACV 2020). For an edge with fundamental matrix F and known principal points pp_i, pp_j, a valid essential matrix E = K_j^T F K_i (with full intrinsics) must have singular values (sigma, sigma, 0). The constructor folds the principal points into G = K_j^T F K_i (focals deferred) and performs a single SVD to precompute the coefficient vectors d_01, d_12; the per-iteration residual is then a cheap closed form in (f_i, f_j) with analytic Jacobians.

Variables are the two focal lengths, each a scalar double. Principal points are constant for the edge and live in the constructor, not in the optimized state.

Inheritance diagram for gtsam::SelfCalibrationFactor:

Public Member Functions

 SelfCalibrationFactor (Key fi_key, Key fj_key, const Matrix3 &F, const Vector2 &pp_i, const Vector2 &pp_j, const SharedNoiseModel &model=nullptr)
Vector2 evaluateError (const double &fi, const double &fj, OptionalMatrixType H1=nullptr, OptionalMatrixType H2=nullptr) const override
 2-vector essential-matrix deviation; Jacobians are 2x1 each.
Public Member Functions inherited from gtsam::NoiseModelFactorT< Vector2, double, double >
Key key () const
 Returns a key.
Vector unwhitenedError (const Values &x, std::vector< Matrix > &H) const
 support taking in the actual vector instead of the pointer as well to get access to this version of the function from derived classes one will need to use the "using" keyword and specify that like this: public: using NoiseModelFactor::unwhitenedError;
 NoiseModelFactorT ()
 Default Constructor for I/O.
 NoiseModelFactorT (const SharedNoiseModel &noiseModel, KeyType< ValueTypes >... keys)
 Constructor.
 NoiseModelFactorT (const SharedNoiseModel &noiseModel, CONTAINER keys)
 Constructor.
virtual Vector2 evaluateError (const ValueTypes &... x, OptionalMatrixTypeT< ValueTypes >... H) const=0
 Override evaluateError to finish implementing an n-way factor.
Vector evaluateError (const ValueTypes &... x, MatrixTypeT< ValueTypes > &... H) const
 If all the optional arguments are matrices then redirect the call to the one which takes pointers.
Vector evaluateError (const ValueTypes &... x) const
 No-Jacobians requested function overload.
AreAllMatrixRefs< Vector, OptionalJacArgs... > evaluateError (const ValueTypes &... x, OptionalJacArgs &&... H) const
 Some (but not all) optional Jacobians are omitted (function overload) and the jacobians are l-value references to matrices.
AreAllMatrixPtrs< Vector, OptionalJacArgs... > evaluateError (const ValueTypes &... x, OptionalJacArgs &&... H) const
 Some (but not all) optional Jacobians are omitted (function overload) and the jacobians are pointers to matrices.
Key key1 () const
Key key2 () const
Key key3 () const
Key key4 () const
Key key5 () const
Key key6 () const
std::shared_ptr< GaussianFactor > linearize (const Values &values) const override
 Linearize factors whose error and argument dimensions are all fixed to an arbitrary-arity FixedJacobianFactor.
Vector unwhitenedError (const Values &x, OptionalMatrixVecType H=nullptr) const override
 This implements the unwhitenedError virtual function by calling the n-key specific version of evaluateError, which is pure virtual so must be implemented in the derived class.
Public Member Functions inherited from gtsam::NoiseModelFactor
 NoiseModelFactor ()
 Default constructor for I/O only.
 ~NoiseModelFactor () override
 Destructor.
template<typename CONTAINER>
 NoiseModelFactor (const SharedNoiseModel &noiseModel, const CONTAINER &keys)
 Constructor.
void print (const std::string &s="", const KeyFormatter &keyFormatter=DefaultKeyFormatter) const override
 Print.
bool equals (const NonlinearFactor &f, double tol=1e-9) const override
 Check if two factors are equal.
size_t dim () const override
 get the dimension of the factor (number of rows on linearization)
const SharedNoiseModelnoiseModel () const
 access to the noise model
Vector unwhitenedError (const Values &x, std::vector< Matrix > &H) const
 support taking in the actual vector instead of the pointer as well to get access to this version of the function from derived classes one will need to use the "using" keyword and specify that like this: public: using NoiseModelFactor::unwhitenedError;
Vector whitenedError (const Values &c) const
 Vector of errors, whitened This is the raw error, i.e., i.e.
Vector unweightedWhitenedError (const Values &c) const
 Vector of errors, whitened, but unweighted by any loss function.
double weight (const Values &c) const
 Compute the effective weight of the factor from the noise model.
double error (const Values &c) const override
 Calculate the error of the factor.
shared_ptr cloneWithNewNoiseModel (const SharedNoiseModel newNoise) const
 Creates a shared_ptr clone of the factor with a new noise model.
double error (const HybridValues &c) const override
 All factor types need to implement an error function.
 NonlinearFactor ()
 Default constructor for I/O only.
template<typename CONTAINER>
 NonlinearFactor (const CONTAINER &keys)
 Constructor from a collection of the keys involved in this factor.
void print (const std::string &s="", const KeyFormatter &keyFormatter=DefaultKeyFormatter) const override
 print
double error (const HybridValues &c) const override
 All factor types need to implement an error function.
virtual bool active (const Values &c) const
 Checks whether a factor should be used based on a set of values.
virtual void qcqpFactors (NonlinearFactorGraph *costs, NonlinearEqualityConstraints *constraints, size_t columnDimension=1) const
 Add this factor's QCQP cost and constraints over matrix-valued QCQP variables with the given column dimension.
virtual shared_ptr clone () const
 Creates a shared_ptr clone of the factor - needs to be specialized to allow for subclasses.
virtual shared_ptr rekey (const std::map< Key, Key > &rekey_mapping) const
 Creates a shared_ptr clone of the factor with different keys using a map from old->new keys.
virtual shared_ptr rekey (const KeyVector &new_keys) const
 Clones a factor and fully replaces its keys.
virtual bool sendable () const
 Should the factor be evaluated in the same thread as the caller This is to enable factors that has shared states (like the Python GIL lock).
Public Member Functions inherited from gtsam::Factor
virtual ~Factor ()=default
 Default destructor.
bool empty () const
 Whether the factor is empty (involves zero variables).
Key front () const
 First key.
Key back () const
 Last key.
const_iterator find (Key key) const
 find
const KeyVectorkeys () const
 Access the factor's involved variable keys.
const_iterator begin () const
 Iterator at beginning of involved variable keys.
const_iterator end () const
 Iterator at end of involved variable keys.
size_t size () const
virtual void printKeys (const std::string &s="Factor", const KeyFormatter &formatter=DefaultKeyFormatter) const
 print only keys
bool equals (const This &other, double tol=1e-9) const
 check equality
KeyVectorkeys ()
iterator begin ()
 Iterator at beginning of involved variable keys.
iterator end ()
 Iterator at end of involved variable keys.

Public Types

using Base = NoiseModelFactorT<Vector2, double, double>
using This = SelfCalibrationFactor
Public Types inherited from gtsam::NoiseModelFactorT< Vector2, double, double >
using ValueType
 The type of the I'th template param can be obtained as ValueType.
Public Types inherited from gtsam::NoiseModelFactor
typedef std::shared_ptr< This > shared_ptr
 Noise model.
Public Types inherited from gtsam::NonlinearFactor
typedef std::shared_ptr< This > shared_ptr
Public Types inherited from gtsam::Factor
typedef KeyVector::iterator iterator
 Iterator over keys.
typedef KeyVector::const_iterator const_iterator
 Const iterator over keys.

Additional Inherited Members

Static Public Attributes inherited from gtsam::NoiseModelFactorT< Vector2, double, double >
static constexpr auto N
 N is the number of variables (N-way factor).
Protected Types inherited from gtsam::NoiseModelFactorT< Vector2, double, double >
using Base
using This
using OptionalMatrixTypeT
using KeyType
using MatrixTypeT
using IsConvertible
using IndexIsValid
using ContainerElementType
using IsContainerOfKeys
using AreAllMatrixRefs
 A helper alias to check if a list of args are all references to a matrix or not.
using IsMatrixPointer
using IsNullpointer
using AreAllMatrixPtrs
 A helper alias to check if a list of args are all pointers to a matrix or not.
Protected Types inherited from gtsam::NoiseModelFactor
typedef NonlinearFactor Base
typedef NoiseModelFactor This
Protected Types inherited from gtsam::NonlinearFactor
typedef Factor Base
typedef NonlinearFactor This
Protected Member Functions inherited from gtsam::NoiseModelFactor
 NoiseModelFactor (const SharedNoiseModel &noiseModel)
 Constructor - only for subclasses, as this does not set keys.
 Factor ()
 Default constructor for I/O.
template<typename CONTAINER>
 Factor (const CONTAINER &keys)
 Construct factor from container of keys.
template<typename ITERATOR>
 Factor (ITERATOR first, ITERATOR last)
 Construct factor from iterator keys.
template<typename CONTAINER>
static Factor FromKeys (const CONTAINER &keys)
 Construct factor from container of keys.
template<typename ITERATOR>
static Factor FromIterators (ITERATOR first, ITERATOR last)
 Construct factor from iterator keys.
Protected Attributes inherited from gtsam::NoiseModelFactor
SharedNoiseModel noiseModel_
Protected Attributes inherited from gtsam::Factor
KeyVector keys_
 The keys involved in this factor.

Constructor & Destructor Documentation

◆ SelfCalibrationFactor()

gtsam::SelfCalibrationFactor::SelfCalibrationFactor ( Key fi_key,
Key fj_key,
const Matrix3 & F,
const Vector2 & pp_i,
const Vector2 & pp_j,
const SharedNoiseModel & model = nullptr )
inline
Parameters
fi_keyKey of the focal length of camera i (F's right index).
fj_keyKey of the focal length of camera j (F's left index).
F3x3 fundamental matrix with convention x_j^T F x_i = 0.
pp_iPrincipal point of camera i.
pp_jPrincipal point of camera j.
modelOptional 2-dimensional noise model (e.g. robust Cauchy); defaults to unit noise when null.

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