Class AstrometryTransform

Inheritance Relationships

Derived Types

Class Documentation

class AstrometryTransform

a virtual (interface) class for geometric transformations.

We implement here One AstrometryTransform interface class, and actual derived classes. Composition in the usual (mathematical) sense is provided using compose(), and some classes (e.g. AstrometryTransformLinear

) handle a * operator. Generic inversion by iteration exists, but it is at least 10 times slower than the corresponding “direct

transformation”. If a transform has an analytical inverse, then providing inverseTransform is obviously a very good idea. Before resorting to inverseTransform, consider using

StarMatchList::inverseTransform(). AstrometryTransformLinear::inverted() and TanPixelToRaDec::inverted() exist. The classes also provide derivation and linear approximation.

Subclassed by lsst::jointcal::AstrometryTransformIdentity, lsst::jointcal::AstrometryTransformPolynomial, lsst::jointcal::AstrometryTransformSkyWcs, lsst::jointcal::BaseTanWcs, lsst::jointcal::TanRaDecToPixel, lsst::jointcal::UserTransform

Public Functions

virtual void apply(const double xIn, const double yIn, double &xOut, double &yOut) const = 0
void apply(Point const &in, Point &out) const

applies the tranfo to in and writes into out. Is indeed virtual.

Point apply(Point const &in) const

All these apply(..) shadow the virtual one in derived classes, unless one writes “using

AstrometryTransform::apply”.

Frame apply(Frame const &inputframe, bool inscribed) const

Transform a bounding box, taking either the inscribed or circumscribed box.

Return

The transformed frame.

Parameters

  • [in] inputframe: The frame to be transformed.

  • [in] inscribed: Return the inscribed (true) or circumscribed (false) box.

virtual void dump(std::ostream &stream = std::cout) const = 0

dumps the transform coefficients to stream.

std::string __str__()
virtual double fit(StarMatchList const &starMatchList) = 0

fits a transform to a std::list of Point pairs (p1,p2, the Point fields in StarMatch).

After the fit this(p1) yields approximately p2. The returned value is the sum of squared residuals. If you want to fit a partial transform (e.g. such that this(T1(p1)) = T2(p2), use StarMatchList::applyTransform beforehand.

void transformStar(FatPoint &in) const
virtual double getJacobian(Point const &point) const

returns the local jacobian.

virtual std::unique_ptr<AstrometryTransform> clone() const = 0

returns a copy (allocated by new) of the transformation.

virtual std::unique_ptr<AstrometryTransform> composeAndReduce(AstrometryTransform const &right) const

Return a reduced composition of newTransform = this(right()), or nullptr if it cannot be reduced.

“Reduced” in this context means that they are capable of being merged into a single transform, for example, for two polynomials:

\[ f(x) = 1 + x^2, g(x) = -1 + 3x \]
we would have h = f.composeAndReduce(g) == 2 - 6x + 9x^2.

To be overloaded by derived classes if they can properly reduce the composition.

Return

The new reduced and composed AstrometryTransform, or nullptr if no such reduction is possible.

Parameters

  • right: The transform to apply first.

virtual double getJacobian(const double x, const double y) const

returns the local jacobian.

virtual void computeDerivative(Point const &where, AstrometryTransformLinear &derivative, const double step = 0.01) const

Computes the local Derivative of a transform, w.r.t. position.

Step is used for numerical derivation.

virtual AstrometryTransformLinear linearApproximation(Point const &where, const double step = 0.01) const

linear (local) approximation.

virtual void transformPosAndErrors(const FatPoint &in, FatPoint &out) const
virtual void transformErrors(Point const &where, const double *vIn, double *vOut) const

transform errors (represented as double[3] in order V(xx),V(yy),Cov(xy))

virtual std::unique_ptr<AstrometryTransform> inverseTransform(const double precision, const Frame &region) const

returns an inverse transform. Numerical if not overloaded.

precision and region refer to the “input” side of this, and hence to the output side of the returned AstrometryTransform.

void getParams(double *params) const

params should be at least Npar() long

void offsetParams(Eigen::VectorXd const &delta)
virtual double paramRef(Eigen::Index const i) const
virtual double &paramRef(Eigen::Index const i)
virtual void paramDerivatives(Point const &where, double *dx, double *dy) const

Derivative w.r.t parameters. Derivatives should be al least 2*NPar long. first Npar, for x, last Npar for y.

virtual std::unique_ptr<AstrometryTransform> roughInverse(const Frame &region) const

Rough inverse.

Stored by the numerical inverter to guess starting point for the trials. Just here to enable overloading.

virtual std::size_t getNpar() const

returns the number of parameters (to compute chi2’s)

virtual std::shared_ptr<ast::Mapping> toAstMap(jointcal::Frame const &domain) const

Create an equivalent AST mapping for this transformation, including an analytic inverse if possible.

Return

An AST Mapping that represents this transformation.

Parameters

  • domain: The domain of the transform, to help find an inverse.

void write(const std::string &fileName) const
virtual void write(std::ostream &stream) const
virtual ~AstrometryTransform()