astrometry.py

Go to the documentation of this file.

#
# LSST Data Management System
# Copyright 2008, 2009, 2010, 2011 LSST Corporation.
#
# This product includes software developed by the
# LSST Project (http://www.lsst.org/).
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the LSST License Statement and
# the GNU General Public License along with this program.  If not,
# see <http://www.lsstcorp.org/LegalNotices/>.
#
import numpy
from contextlib import contextmanager

import lsst.pex.exceptions
import lsst.afw.geom as afwGeom
from lsst.afw.cameraGeom import TAN_PIXELS
import lsst.pex.config as pexConfig
import lsst.pipe.base as pipeBase
from lsst.meas.astrom.astrom import Astrometry
from lsst.meas.astrom.sip import makeCreateWcsWithSip
from lsst.afw.table import Point2DKey, CovarianceMatrix2fKey

class AstrometryConfig(pexConfig.Config):
    solver = pexConfig.ConfigField(
        dtype = Astrometry.ConfigClass,
        doc = "Configuration for the astrometry solver",
    )
    forceKnownWcs = pexConfig.Field(dtype=bool, doc=(
        "Assume that the input image's WCS is correct, without comparing it to any external reality." +
        " (In contrast to using Astrometry.net).  NOTE, if you set this, you probably also want to" +
        " un-set 'solver.calculateSip'; otherwise we'll still try to find a TAN-SIP WCS starting " +
        " from the existing WCS"), default=False)
    rejectThresh = pexConfig.RangeField(dtype=float, default=3.0, doc="Rejection threshold for Wcs fitting",
                                        min=0.0, inclusiveMin=False)
    rejectIter = pexConfig.RangeField(dtype=int, default=3, doc="Rejection iterations for Wcs fitting",
                                      min=0)


## \addtogroup LSST_task_documentation
## \{
## \page AstrometryTask
## \ref AstrometryTask_ "AstrometryTask"
## \copybrief AstrometryTask
## \}

class AstrometryTask(pipeBase.Task):
    """!
\anchor AstrometryTask_

\brief %Match input sources with a reference catalog and solve for the Wcs

\warning This task is currently in \link lsst.pipe.tasks.astrometry\endlink;
it should be moved to lsst.meas.astrom.

The actual matching and solving is done by the 'solver'; this Task
serves as a wrapper for taking into account the known optical distortion.

\section pipe_tasks_astrometry_Contents Contents

 - \ref pipe_tasks_astrometry_Purpose
 - \ref pipe_tasks_astrometry_Initialize
 - \ref pipe_tasks_astrometry_IO
 - \ref pipe_tasks_astrometry_Config
 - \ref pipe_tasks_astrometry_Debug
 - \ref pipe_tasks_astrometry_Example

\section pipe_tasks_astrometry_Purpose  Description

\copybrief AstrometryTask

Calculate an Exposure's zero-point given a set of flux measurements of stars matched to an input catalogue.
The type of flux to use is specified by AstrometryConfig.fluxField.

The algorithm clips outliers iteratively, with parameters set in the configuration.

\note This task can adds fields to the schema, so any code calling this task must ensure that
these columns are indeed present in the input match list; see \ref pipe_tasks_astrometry_Example

\section pipe_tasks_astrometry_Initialize       Task initialisation

\copydoc \_\_init\_\_

\section pipe_tasks_astrometry_IO               Invoking the Task

\copydoc run

\section pipe_tasks_astrometry_Config       Configuration parameters

See \ref AstrometryConfig

\section pipe_tasks_astrometry_Debug            Debug variables

The \link lsst.pipe.base.cmdLineTask.CmdLineTask command line task\endlink interface supports a
flag \c -d to import \b debug.py from your \c PYTHONPATH; see \ref baseDebug for more about \b debug.py files.

The available variables in AstrometryTask are:
<DL>
  <DT> \c display
  <DD> If True call astrometry.showAstrometry while iterating AstrometryConfig.rejectIter times,
  and also after converging.
  <DT> \c frame
  <DD> ds9 frame to use in astrometry.showAstrometry
  <DT> \c pause
  <DD> Pause within astrometry.showAstrometry
</DL>

\section pipe_tasks_astrometry_Example  A complete example of using AstrometryTask

See \ref meas_photocal_photocal_Example.

To investigate the \ref pipe_tasks_astrometry_Debug, put something like
\code{.py}
    import lsstDebug
    def DebugInfo(name):
        di = lsstDebug.getInfo(name)        # N.b. lsstDebug.Info(name) would call us recursively
        if name == "lsst.pipe.tasks.astrometry":
            di.display = 1

        return di

    lsstDebug.Info = DebugInfo
\endcode
into your debug.py file and run photoCalTask.py with the \c --debug flag.
    """
    ConfigClass = AstrometryConfig

    # Need init as well as __init__ because "\copydoc __init__" fails (doxygen bug 732264)
    def __init__(self, schema, **kwds):
        """!Create the astrometric calibration task.  Most arguments are simply passed onto pipe.base.Task.

        \param schema An lsst::afw::table::Schema used to create the output lsst.afw.table.SourceCatalog
        \param **kwds keyword arguments to be passed to the lsst.pipe.base.task.Task constructor

        A centroid field "centroid.distorted" (used internally during the Task's operation)
        will be added to the schema.
        """
        pipeBase.Task.__init__(self, **kwds)

        if schema.getVersion() == 0:
            self.distortedName = "centroid.distorted"
            self.centroidKey = schema.addField(self.distortedName, type="PointD",
                                           doc="centroid distorted for astrometry solver")
            self.centroidErrKey = schema.addField(self.distortedName + ".err", type="CovPointF",
                                           doc="centroid distorted err for astrometry solver")
            self.centroidFlagKey = schema.addField(self.distortedName + ".flag", type="Flag",
                                           doc="centroid distorted flag astrometry solver")
        else:
            self.distortedName = "astrom_distorted"
            self.centroidXKey = schema.addField(self.distortedName + "_x", type="D",
                                           doc="centroid distorted for astrometry solver")
            self.centroidYKey = schema.addField(self.distortedName + "_y", type="D",
                                           doc="centroid distorted for astrometry solver")
            self.centroidXErrKey = schema.addField(self.distortedName + "_xSigma", type="F",
                                           doc="centroid distorted err for astrometry solver")
            self.centroidYErrKey = schema.addField(self.distortedName + "_ySigma", type="F",
                                           doc="centroid distorted err for astrometry solver")
            self.centroidFlagKey = schema.addField(self.distortedName + "_flag", type="Flag",
                                           doc="centroid distorted flag astrometry solver")
            self.centroidKey = Point2DKey(self.centroidXKey, self.centroidYKey)
            self.centroidErrKey = CovarianceMatrix2fKey((self.centroidXErrKey, self.centroidYErrKey))
 
        self.astrometer = None

    @pipeBase.timeMethod
    def run(self, exposure, sources):
        """!Match with reference sources and calculate an astrometric solution

        \param exposure Exposure to calibrate
        \param sources SourceCatalog of measured sources
        \return a pipeBase.Struct with fields:
        - matches: Astrometric matches
        - matchMeta: Metadata for astrometric matches

        The reference catalog actually used is up to the implementation
        of the solver; it will be manifested in the returned matches as
        a list of lsst.afw.table.ReferenceMatch objects (\em i.e. of lsst.afw.table.Match with
        \c first being of type lsst.afw.table.SimpleRecord and \c second type lsst.afw.table.SourceRecord ---
        the reference object and matched object respectively).

        \note
        The input sources have the centroid slot moved to a new column "centroid.distorted"
        which has the positions corrected for any known optical distortion;
        the 'solver' (which is instantiated in the 'astrometry' member)
        should therefore simply use the centroids provided by calling
        afw.table.Source.getCentroid() on the individual source records.  This column \em must
        be present in the sources table.
        """
        with self.distortionContext(exposure, sources) as bbox:
            results = self.astrometry(exposure, sources, bbox=bbox)

        if results.matches:
            self.refitWcs(exposure, sources, results.matches)

        return results

    @pipeBase.timeMethod
    def distort(self, exposure, sources):
        """!Calculate distorted source positions

        CCD images are often affected by optical distortion that makes
        the astrometric solution higher order than linear.  Unfortunately,
        most (all?) matching algorithms require that the distortion be
        small or zero, and so it must be removed.  We do this by calculating
        (un-)distorted positions, based on a known optical distortion model
        in the Ccd.

        The distortion correction moves sources, so we return the distorted
        bounding box.

        \param[in]     exposure Exposure to process
        \param[in,out] sources  SourceCatalog; getX() and getY() will be used as inputs,
                                with distorted points in "centroid.distorted" field.
        \return bounding box of distorted exposure
        """
        detector = exposure.getDetector()
        pixToTanXYTransform = None
        if detector is None:
            self.log.warn("No detector associated with exposure; assuming null distortion")
        else:
            tanSys = detector.makeCameraSys(TAN_PIXELS)
            pixToTanXYTransform = detector.getTransformMap().get(tanSys)

        if pixToTanXYTransform is None:
            self.log.info("Null distortion correction")
            for s in sources:
                s.set(self.centroidKey, s.getCentroid())
                s.set(self.centroidErrKey, s.getCentroidErr())
                s.set(self.centroidFlagKey, s.getCentroidFlag())
            return exposure.getBBox()

        # Distort source positions
        self.log.info("Applying distortion correction")
        for s in sources:
            centroid = pixToTanXYTransform.forwardTransform(s.getCentroid())
            s.set(self.centroidKey, centroid)
            s.set(self.centroidErrKey, s.getCentroidErr())
            s.set(self.centroidFlagKey, s.getCentroidFlag())


        # Get distorted image size so that astrometry_net does not clip.
        bboxD = afwGeom.Box2D()
        for corner in detector.getCorners(TAN_PIXELS):
            bboxD.include(corner)
        return afwGeom.Box2I(bboxD)

    @contextmanager
    def distortionContext(self, exposure, sources):
        """!Context manager that applies and removes distortion

        We move the "centroid" definition in the catalog table to
        point to the distorted positions.  This is undone on exit
        from the context.

        The input Wcs is taken to refer to the coordinate system
        with the distortion correction applied, and hence no shift
        is required when the sources are distorted.  However, after
        Wcs fitting, the Wcs is in the distorted frame so when the
        distortion correction is removed, the Wcs needs to be
        shifted to compensate.

        \param exposure Exposure holding Wcs
        \param sources Sources to correct for distortion
        \return bounding box of distorted exposure
        """
        # Apply distortion
        bbox = self.distort(exposure, sources)
        oldCentroidName = sources.table.getCentroidDefinition()
        sources.table.defineCentroid(self.distortedName)
        try:
            yield bbox # Execute 'with' block, providing bbox to 'as' variable
        finally:
            # Un-apply distortion
            sources.table.defineCentroid(oldCentroidName)
            x0, y0 = exposure.getXY0()
            wcs = exposure.getWcs()
            if wcs:
                wcs.shiftReferencePixel(-bbox.getMinX() + x0, -bbox.getMinY() + y0)

    @pipeBase.timeMethod
    def astrometry(self, exposure, sources, bbox=None):
        """!Solve astrometry to produce WCS

        \param exposure Exposure to process
        \param sources Sources
        \param bbox Bounding box, or None to use exposure
        \return a pipe.base.Struct with fields:
         - matches: star matches
         - matchMeta: match metadata
        """
        if not self.config.forceKnownWcs:
            self.log.info("Solving astrometry")
        if bbox is None:
            bbox = exposure.getBBox()

        if not self.astrometer:
            self.astrometer = Astrometry(self.config.solver, log=self.log)

        kwargs = dict(x0=bbox.getMinX(), y0=bbox.getMinY(), imageSize=bbox.getDimensions())
        if self.config.forceKnownWcs:
            self.log.info("Forcing the input exposure's WCS")
            if self.config.solver.calculateSip:
                self.log.warn("'forceKnownWcs' and 'solver.calculateSip' options are both set." +
                              " Will try to compute a TAN-SIP WCS starting from the input WCS.")
            astrom = self.astrometer.useKnownWcs(sources, exposure=exposure, **kwargs)
        else:
            astrom = self.astrometer.determineWcs(sources, exposure, **kwargs)

        if astrom is None or astrom.getWcs() is None:
            raise RuntimeError("Unable to solve astrometry")

        matches = astrom.getMatches()
        matchMeta = astrom.getMatchMetadata()
        if matches is None or len(matches) == 0:
            raise RuntimeError("No astrometric matches")
        self.log.info("%d astrometric matches" %  (len(matches)))

        if not self.config.forceKnownWcs:
            # Note that this is the Wcs for the provided positions, which may be distorted
            exposure.setWcs(astrom.getWcs())

        self.display('astrometry', exposure=exposure, sources=sources, matches=matches)

        return pipeBase.Struct(matches=matches, matchMeta=matchMeta)

    @pipeBase.timeMethod
    def refitWcs(self, exposure, sources, matches):
        """!A final Wcs solution after matching and removing distortion

        Specifically, fitting the non-linear part, since the linear
        part has been provided by the matching engine.

        \param exposure Exposure of interest
        \param sources Sources on image (no distortion applied)
        \param matches Astrometric matches

        \return the resolved-Wcs object, or None if config.solver.calculateSip is False.
        """
        sip = None
        if self.config.solver.calculateSip:
            self.log.info("Refitting WCS")
            origMatches = matches
            wcs = exposure.getWcs()

            import lsstDebug
            display = lsstDebug.Info(__name__).display
            frame = lsstDebug.Info(__name__).frame
            pause = lsstDebug.Info(__name__).pause

            def fitWcs(initialWcs, title=None):
                """!Do the WCS fitting and display of the results"""
                sip = makeCreateWcsWithSip(matches, initialWcs, self.config.solver.sipOrder)
                resultWcs = sip.getNewWcs()
                if display:
                    showAstrometry(exposure, resultWcs, origMatches, matches, frame=frame,
                                   title=title, pause=pause)
                return resultWcs, sip.getScatterOnSky()

            numRejected = 0
            try:
                for i in range(self.config.rejectIter):
                    wcs, scatter = fitWcs(wcs, title="Iteration %d" % i)

                    ref = numpy.array([wcs.skyToPixel(m.first.getCoord()) for m in matches])
                    src = numpy.array([m.second.getCentroid() for m in matches])
                    diff = ref - src
                    rms = diff.std()
                    trimmed = []
                    for d, m in zip(diff, matches):
                        if numpy.all(numpy.abs(d) < self.config.rejectThresh*rms):
                            trimmed.append(m)
                        else:
                            numRejected += 1
                    if len(matches) == len(trimmed):
                        break
                    matches = trimmed

                # Final fit after rejection iterations
                wcs, scatter = fitWcs(wcs, title="Final astrometry")

            except lsst.pex.exceptions.LengthError as e:
                self.log.warn("Unable to fit SIP: %s" % e)

            self.log.info("Astrometric scatter: %f arcsec (%s non-linear terms, %d matches, %d rejected)" %
                          (scatter.asArcseconds(), "with" if wcs.hasDistortion() else "without",
                           len(matches), numRejected))
            exposure.setWcs(wcs)

            # Apply WCS to sources
            for index, source in enumerate(sources):
                sky = wcs.pixelToSky(source.getX(), source.getY())
                source.setCoord(sky)
        else:
            self.log.warn("Not calculating a SIP solution; matches may be suspect")

        self.display('astrometry', exposure=exposure, sources=sources, matches=matches)

        return sip


def showAstrometry(exposure, wcs, allMatches, useMatches, frame=0, title=None, pause=False):
    """!Show results of astrometry fitting

    \param exposure Image to display
    \param wcs Astrometric solution
    \param allMatches List of all astrometric matches (including rejects)
    \param useMatches List of used astrometric matches
    \param frame Frame number for display
    \param title Title for display
    \param pause Pause to allow viewing of the display and optional debugging?

    - Matches are shown in yellow if used in the Wcs solution, otherwise red
     - +: Detected objects
     - x: Catalogue objects
    """
    import lsst.afw.display.ds9 as ds9
    ds9.mtv(exposure, frame=frame, title=title)

    useIndices = set(m.second.getId() for m in useMatches)

    radii = []
    with ds9.Buffering():
        for i, m in enumerate(allMatches):
            x, y = m.second.getX(), m.second.getY()
            pix = wcs.skyToPixel(m.first.getCoord())

            isUsed = m.second.getId() in useIndices
            if isUsed:
                radii.append(numpy.hypot(pix[0] - x, pix[1] - y))

            color = ds9.YELLOW if isUsed else ds9.RED

            ds9.dot("+", x, y, size=10, frame=frame, ctype=color)
            ds9.dot("x", pix[0], pix[1], size=10, frame=frame, ctype=color)

    radii = numpy.array(radii)
    print "<dr> = %.4g +- %.4g pixels [%d/%d matches]" % (radii.mean(), radii.std(),
                                                          len(useMatches), len(allMatches))

    if pause:
        import sys
        while True:
            try:
                reply = raw_input("Debugging? [p]db [q]uit; any other key to continue... ").strip()
            except EOFError:
                reply = ""

            reply = reply.split()
            if len(reply) > 1:
                reply = reply[0]
            if reply == "p":
                import pdb;pdb.set_trace()
            elif reply == "q":
                sys.exit(1)
            else:
                break