Coverage for python/lsst/meas/algorithms/detection.py: 15%
346 statements
« prev ^ index » next coverage.py v7.14.1, created at 2026-05-30 01:59 -0700
« prev ^ index » next coverage.py v7.14.1, created at 2026-05-30 01:59 -0700
1#
2# LSST Data Management System
3#
4# Copyright 2008-2017 AURA/LSST.
5#
6# This product includes software developed by the
7# LSST Project (http://www.lsst.org/).
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the LSST License Statement and
20# the GNU General Public License along with this program. If not,
21# see <https://www.lsstcorp.org/LegalNotices/>.
22#
24__all__ = ("PsfGenerationError", "SourceDetectionConfig", "SourceDetectionTask", "addExposures")
26from contextlib import contextmanager
28import numpy as np
30import lsst.geom
31import lsst.afw.display as afwDisplay
32import lsst.afw.detection as afwDet
33import lsst.afw.geom as afwGeom
34import lsst.afw.image as afwImage
35import lsst.afw.math as afwMath
36import lsst.afw.table as afwTable
37import lsst.pex.config as pexConfig
38import lsst.pipe.base as pipeBase
39from lsst.utils.timer import timeMethod
40from .subtractBackground import SubtractBackgroundTask, backgroundFlatContext
43class PsfGenerationError(pipeBase.AlgorithmError):
44 """Raised when we cannot generate a PSF for detection
46 Parameters
47 ----------
48 msg : `str`
49 Error message.
50 **kwargs : `dict`, optional
51 Additional keyword arguments to initialize the Exception base class.
52 """
53 def __init__(self, msg, **kwargs):
54 self.msg = msg
55 self._metadata = kwargs
56 super().__init__(msg, **kwargs)
58 def __str__(self):
59 # Exception doesn't handle **kwargs, so we need a custom str.
60 return f"{self.msg}: {self.metadata}"
62 @property
63 def metadata(self):
64 for key, value in self._metadata.items():
65 if not isinstance(value, (int, float, str)):
66 raise TypeError(f"{key} is of type {type(value)}, but only (int, float, str) are allowed.")
67 return self._metadata
70class SourceDetectionConfig(pexConfig.Config):
71 """Configuration parameters for the SourceDetectionTask
72 """
73 minPixels = pexConfig.RangeField(
74 doc="detected sources with fewer than the specified number of pixels will be ignored",
75 dtype=int, optional=False, default=1, min=0,
76 )
77 isotropicGrow = pexConfig.Field(
78 doc="Grow pixels as isotropically as possible? If False, use a Manhattan metric instead.",
79 dtype=bool, default=True,
80 )
81 combinedGrow = pexConfig.Field(
82 doc="Grow all footprints at the same time? This allows disconnected footprints to merge.",
83 dtype=bool, default=True,
84 )
85 nSigmaToGrow = pexConfig.Field(
86 doc="Grow detections by nSigmaToGrow * [PSF RMS width]; if 0 then do not grow",
87 dtype=float, default=2.4, # 2.4 pixels/sigma is roughly one pixel/FWHM
88 )
89 returnOriginalFootprints = pexConfig.Field(
90 doc="Grow detections to set the image mask bits, but return the original (not-grown) footprints",
91 dtype=bool, optional=False, default=False,
92 )
93 thresholdValue = pexConfig.RangeField(
94 doc="Threshold for detecting footprints; exact meaning and units depend on thresholdType.",
95 dtype=float, optional=False, default=5.0, min=0.0,
96 )
97 includeThresholdMultiplier = pexConfig.RangeField(
98 doc="Multiplier on thresholdValue for whether a source is included in the output catalog."
99 " For example, thresholdValue=5, includeThresholdMultiplier=10, thresholdType='pixel_stdev' "
100 "results in a catalog of sources at >50 sigma with the detection mask and footprints "
101 "including pixels >5 sigma.",
102 dtype=float, default=1.0, min=0.0,
103 )
104 thresholdType = pexConfig.ChoiceField(
105 doc="Specifies the meaning of thresholdValue.",
106 dtype=str, optional=False, default="pixel_stdev",
107 allowed={
108 "variance": "threshold applied to image variance",
109 "stdev": "threshold applied to image std deviation",
110 "value": "threshold applied to image value",
111 "pixel_stdev": "threshold applied to per-pixel std deviation",
112 },
113 )
114 thresholdPolarity = pexConfig.ChoiceField(
115 doc="Specifies whether to detect positive, or negative sources, or both.",
116 dtype=str, optional=False, default="positive",
117 allowed={
118 "positive": "detect only positive sources",
119 "negative": "detect only negative sources",
120 "both": "detect both positive and negative sources",
121 },
122 )
123 adjustBackground = pexConfig.Field(
124 dtype=float,
125 doc="Fiddle factor to add to the background; debugging only",
126 default=0.0,
127 )
128 reEstimateBackground = pexConfig.Field(
129 dtype=bool,
130 doc="Estimate the background again after final source detection?",
131 default=True, optional=False,
132 )
133 doApplyFlatBackgroundRatio = pexConfig.Field(
134 doc="Convert from a photometrically flat image to one suitable for background subtraction? "
135 "Only used if reEstimateBackground is True."
136 "If True, then a backgroundToPhotometricRatio must be supplied to the task run method.",
137 dtype=bool,
138 default=False,
139 )
140 background = pexConfig.ConfigurableField(
141 doc="Background re-estimation; ignored if reEstimateBackground false",
142 target=SubtractBackgroundTask,
143 )
144 tempLocalBackground = pexConfig.ConfigurableField(
145 doc=("A local (small-scale), temporary background estimation step run between "
146 "detecting above-threshold regions and detecting the peaks within "
147 "them; used to avoid detecting spuerious peaks in the wings."),
148 target=SubtractBackgroundTask,
149 )
150 doTempLocalBackground = pexConfig.Field(
151 dtype=bool,
152 doc="Enable temporary local background subtraction? (see tempLocalBackground)",
153 default=True,
154 )
155 tempWideBackground = pexConfig.ConfigurableField(
156 doc=("A wide (large-scale) background estimation and removal before footprint and peak detection. "
157 "It is added back into the image after detection. The purpose is to suppress very large "
158 "footprints (e.g., from large artifacts) that the deblender may choke on."),
159 target=SubtractBackgroundTask,
160 )
161 doTempWideBackground = pexConfig.Field(
162 dtype=bool,
163 doc="Do temporary wide (large-scale) background subtraction before footprint detection?",
164 default=False,
165 )
166 nPeaksMaxSimple = pexConfig.Field(
167 dtype=int,
168 doc=("The maximum number of peaks in a Footprint before trying to "
169 "replace its peaks using the temporary local background"),
170 default=1,
171 )
172 nSigmaForKernel = pexConfig.Field(
173 dtype=float,
174 doc=("Multiple of PSF RMS size to use for convolution kernel bounding box size; "
175 "note that this is not a half-size. The size will be rounded up to the nearest odd integer"),
176 default=7.0,
177 )
178 statsMask = pexConfig.ListField(
179 dtype=str,
180 doc="Mask planes to ignore when calculating statistics of image (for thresholdType=stdev)",
181 default=['BAD', 'SAT', 'EDGE', 'NO_DATA'],
182 )
183 excludeMaskPlanes = lsst.pex.config.ListField(
184 dtype=str,
185 default=[],
186 doc="Mask planes to exclude when detecting sources."
187 )
189 def setDefaults(self):
190 self.tempLocalBackground.binSize = 64
191 self.tempLocalBackground.algorithm = "AKIMA_SPLINE"
192 self.tempLocalBackground.useApprox = False
193 # Background subtraction to remove a large-scale background (e.g., scattered light); restored later.
194 # Want to keep it from exceeding the deblender size limit of 1 Mpix, so half that is reasonable.
195 self.tempWideBackground.binSize = 512
196 self.tempWideBackground.algorithm = "AKIMA_SPLINE"
197 self.tempWideBackground.useApprox = False
198 # Ensure we can remove even bright scattered light that is DETECTED
199 for maskPlane in ("DETECTED", "DETECTED_NEGATIVE"):
200 if maskPlane in self.tempWideBackground.ignoredPixelMask:
201 self.tempWideBackground.ignoredPixelMask.remove(maskPlane)
204class SourceDetectionTask(pipeBase.Task):
205 """Detect peaks and footprints of sources in an image.
207 This task expects the image to have been background subtracted first.
208 Running detection on images with a non-zero-centered background may result
209 in a single source detected on the entire image containing thousands of
210 peaks, or other pathological outputs.
212 Parameters
213 ----------
214 schema : `lsst.afw.table.Schema`
215 Schema object used to create the output `lsst.afw.table.SourceCatalog`
216 **kwds
217 Keyword arguments passed to `lsst.pipe.base.Task.__init__`
219 If schema is not None and configured for 'both' detections,
220 a 'flags.negative' field will be added to label detections made with a
221 negative threshold.
223 Notes
224 -----
225 This task convolves the image with a Gaussian approximation to the PSF,
226 matched to the sigma of the input exposure, because this is separable and
227 fast. The PSF would have to be very non-Gaussian or non-circular for this
228 approximation to have a significant impact on the signal-to-noise of the
229 detected sources.
230 """
231 ConfigClass = SourceDetectionConfig
232 _DefaultName = "sourceDetection"
234 def __init__(self, schema=None, **kwds):
235 pipeBase.Task.__init__(self, **kwds)
236 if schema is not None and self.config.thresholdPolarity == "both":
237 self.negativeFlagKey = schema.addField(
238 "is_negative", type="Flag",
239 doc="Set if source peak was detected as negative."
240 )
241 else:
242 if self.config.thresholdPolarity == "both":
243 self.log.warning("Detection polarity set to 'both', but no flag will be "
244 "set to distinguish between positive and negative detections")
245 self.negativeFlagKey = None
246 if self.config.reEstimateBackground:
247 self.makeSubtask("background")
248 if self.config.doTempLocalBackground:
249 self.makeSubtask("tempLocalBackground")
250 if self.config.doTempWideBackground:
251 self.makeSubtask("tempWideBackground")
253 @timeMethod
254 def run(self, table, exposure, doSmooth=True, sigma=None, clearMask=True, expId=None,
255 background=None, backgroundToPhotometricRatio=None):
256 r"""Detect sources and return catalog(s) of detections.
258 Parameters
259 ----------
260 table : `lsst.afw.table.SourceTable`
261 Table object that will be used to create the SourceCatalog.
262 exposure : `lsst.afw.image.Exposure`
263 Exposure to process; DETECTED mask plane will be set in-place.
264 doSmooth : `bool`, optional
265 If True, smooth the image before detection using a Gaussian of width
266 ``sigma``, or the measured PSF width. Set to False when running on
267 e.g. a pre-convolved image, or a mask plane.
268 sigma : `float`, optional
269 Sigma of PSF (pixels); used for smoothing and to grow detections;
270 if None then measure the sigma of the PSF of the exposure
271 clearMask : `bool`, optional
272 Clear DETECTED{,_NEGATIVE} planes before running detection.
273 expId : `int`, optional
274 Exposure identifier; unused by this implementation, but used for
275 RNG seed by subclasses.
276 background : `lsst.afw.math.BackgroundList`, optional
277 Background that was already subtracted from the exposure; will be
278 modified in-place if ``reEstimateBackground=True``.
279 backgroundToPhotometricRatio : `lsst.afw.image.Image`, optional
280 Image to convert photometric-flattened image to
281 background-flattened image if ``reEstimateBackground=True`` and
282 exposure has been photometric-flattened.
284 Returns
285 -------
286 result : `lsst.pipe.base.Struct`
287 The `~lsst.pipe.base.Struct` contains:
289 ``sources``
290 Detected sources on the exposure.
291 (`lsst.afw.table.SourceCatalog`)
292 ``positive``
293 Positive polarity footprints.
294 (`lsst.afw.detection.FootprintSet` or `None`)
295 ``negative``
296 Negative polarity footprints.
297 (`lsst.afw.detection.FootprintSet` or `None`)
298 ``numPos``
299 Number of footprints in positive or 0 if detection polarity was
300 negative. (`int`)
301 ``numNeg``
302 Number of footprints in negative or 0 if detection polarity was
303 positive. (`int`)
304 ``background``
305 Re-estimated background. `None` if
306 ``reEstimateBackground==False``.
307 (`lsst.afw.math.BackgroundList`)
308 ``factor``
309 Multiplication factor applied to the configured detection
310 threshold. (`float`)
312 Raises
313 ------
314 ValueError
315 Raised if flags.negative is needed, but isn't in table's schema.
316 lsst.pipe.base.TaskError
317 Raised if sigma=None, doSmooth=True and the exposure has no PSF.
319 Notes
320 -----
321 If you want to avoid dealing with Sources and Tables, you can use
322 `detectFootprints()` to just get the
323 `~lsst.afw.detection.FootprintSet`\s.
324 """
325 if self.negativeFlagKey is not None and self.negativeFlagKey not in table.getSchema():
326 raise ValueError("Table has incorrect Schema")
327 results = self.detectFootprints(exposure=exposure, doSmooth=doSmooth, sigma=sigma,
328 clearMask=clearMask, expId=expId, background=background,
329 backgroundToPhotometricRatio=backgroundToPhotometricRatio)
330 sources = afwTable.SourceCatalog(table)
331 sources.reserve(results.numPos + results.numNeg)
332 if results.negative:
333 results.negative.makeSources(sources)
334 if self.negativeFlagKey:
335 for record in sources:
336 record.set(self.negativeFlagKey, True)
337 if results.positive:
338 results.positive.makeSources(sources)
339 results.sources = sources
340 return results
342 def display(self, exposure, results, convolvedImage=None):
343 """Display detections if so configured
345 Displays the ``exposure`` in frame 0, overlays the detection peaks.
347 Requires that ``lsstDebug`` has been set up correctly, so that
348 ``lsstDebug.Info("lsst.meas.algorithms.detection")`` evaluates `True`.
350 If the ``convolvedImage`` is non-`None` and
351 ``lsstDebug.Info("lsst.meas.algorithms.detection") > 1``, the
352 ``convolvedImage`` will be displayed in frame 1.
354 Parameters
355 ----------
356 exposure : `lsst.afw.image.Exposure`
357 Exposure to display, on which will be plotted the detections.
358 results : `lsst.pipe.base.Struct`
359 Results of the 'detectFootprints' method, containing positive and
360 negative footprints (which contain the peak positions that we will
361 plot). This is a `Struct` with ``positive`` and ``negative``
362 elements that are of type `lsst.afw.detection.FootprintSet`.
363 convolvedImage : `lsst.afw.image.Image`, optional
364 Convolved image used for thresholding.
365 """
366 try:
367 import lsstDebug
368 display = lsstDebug.Info(__name__).display
369 except ImportError:
370 try:
371 display
372 except NameError:
373 display = False
374 if not display:
375 return
377 afwDisplay.setDefaultMaskTransparency(75)
379 disp0 = afwDisplay.Display(frame=0)
380 disp0.mtv(exposure, title="detection")
382 def plotPeaks(fps, ctype):
383 if fps is None:
384 return
385 with disp0.Buffering():
386 for fp in fps.getFootprints():
387 for pp in fp.getPeaks():
388 disp0.dot("+", pp.getFx(), pp.getFy(), ctype=ctype)
389 plotPeaks(results.positive, "yellow")
390 plotPeaks(results.negative, "red")
392 if convolvedImage and display > 1:
393 disp1 = afwDisplay.Display(frame=1)
394 disp1.mtv(convolvedImage, title="PSF smoothed")
396 disp2 = afwDisplay.Display(frame=2)
397 disp2.mtv(afwImage.ImageF(np.sqrt(exposure.variance.array)), title="stddev")
399 def applyTempLocalBackground(self, exposure, middle, results):
400 """Apply a temporary local background subtraction
402 This temporary local background serves to suppress noise fluctuations
403 in the wings of bright objects.
405 Peaks in the footprints will be updated.
407 Parameters
408 ----------
409 exposure : `lsst.afw.image.Exposure`
410 Exposure for which to fit local background.
411 middle : `lsst.afw.image.MaskedImage`
412 Convolved image on which detection will be performed
413 (typically smaller than ``exposure`` because the
414 half-kernel has been removed around the edges).
415 results : `lsst.pipe.base.Struct`
416 Results of the 'detectFootprints' method, containing positive and
417 negative footprints (which contain the peak positions that we will
418 plot). This is a `Struct` with ``positive`` and ``negative``
419 elements that are of type `lsst.afw.detection.FootprintSet`.
420 """
421 # Subtract the local background from the smoothed image. Since we
422 # never use the smoothed again we don't need to worry about adding
423 # it back in.
424 bg = self.tempLocalBackground.fitBackground(exposure.getMaskedImage())
425 bgImage = bg.getImageF(self.tempLocalBackground.config.algorithm,
426 self.tempLocalBackground.config.undersampleStyle)
427 middle -= bgImage.Factory(bgImage, middle.getBBox())
428 if self.config.thresholdPolarity != "negative":
429 results.positiveThreshold = self.makeThreshold(middle, "positive")
430 self.updatePeaks(results.positive, middle, results.positiveThreshold)
431 if self.config.thresholdPolarity != "positive":
432 results.negativeThreshold = self.makeThreshold(middle, "negative")
433 self.updatePeaks(results.negative, middle, results.negativeThreshold)
435 def clearMask(self, mask):
436 """Clear the DETECTED and DETECTED_NEGATIVE mask planes.
438 Removes any previous detection mask in preparation for a new
439 detection pass.
441 Parameters
442 ----------
443 mask : `lsst.afw.image.Mask`
444 Mask to be cleared.
445 """
446 mask &= ~(mask.getPlaneBitMask("DETECTED") | mask.getPlaneBitMask("DETECTED_NEGATIVE"))
448 def calculateKernelSize(self, sigma):
449 """Calculate the size of the smoothing kernel.
451 Uses the ``nSigmaForKernel`` configuration parameter. Note
452 that that is the full width of the kernel bounding box
453 (so a value of 7 means 3.5 sigma on either side of center).
454 The value will be rounded up to the nearest odd integer.
456 Parameters
457 ----------
458 sigma : `float`
459 Gaussian sigma of smoothing kernel.
461 Returns
462 -------
463 size : `int`
464 Size of the smoothing kernel.
465 """
466 return (int(sigma * self.config.nSigmaForKernel + 0.5)//2)*2 + 1 # make sure it is odd
468 def getPsf(self, exposure, sigma=None):
469 """Create a single Gaussian PSF for an exposure.
471 If ``sigma`` is provided, we make a `~lsst.afw.detection.GaussianPsf`
472 with that, otherwise use the sigma from the psf of the ``exposure`` to
473 make the `~lsst.afw.detection.GaussianPsf`.
475 Parameters
476 ----------
477 exposure : `lsst.afw.image.Exposure`
478 Exposure from which to retrieve the PSF.
479 sigma : `float`, optional
480 Gaussian sigma to use if provided.
482 Returns
483 -------
484 psf : `lsst.afw.detection.GaussianPsf`
485 PSF to use for detection.
487 Raises
488 ------
489 RuntimeError
490 Raised if ``sigma`` is not provided and ``exposure`` does not
491 contain a ``Psf`` object.
492 """
493 if sigma is None:
494 psf = exposure.getPsf()
495 if psf is None:
496 raise PsfGenerationError("Unable to determine PSF to use for detection: no sigma provided")
497 sigma = psf.computeShape(psf.getAveragePosition()).getDeterminantRadius()
498 if not np.isfinite(sigma) or sigma <= 0.0:
499 raise PsfGenerationError("Invalid sigma=%s for PSF used in detection" % (sigma,))
500 size = self.calculateKernelSize(sigma)
501 psf = afwDet.GaussianPsf(size, size, sigma)
502 return psf
504 def convolveImage(self, maskedImage, psf, doSmooth=True):
505 """Convolve the image with the PSF.
507 We convolve the image with a Gaussian approximation to the PSF,
508 because this is separable and therefore fast. It's technically a
509 correlation rather than a convolution, but since we use a symmetric
510 Gaussian there's no difference.
512 The convolution can be disabled with ``doSmooth=False``. If we do
513 convolve, we mask the edges as ``EDGE`` and return the convolved image
514 with the edges removed. This is because we can't convolve the edges
515 because the kernel would extend off the image.
517 Parameters
518 ----------
519 maskedImage : `lsst.afw.image.MaskedImage`
520 Image to convolve.
521 psf : `lsst.afw.detection.Psf`
522 PSF to convolve with (actually with a Gaussian approximation
523 to it).
524 doSmooth : `bool`
525 Actually do the convolution? Set to False when running on
526 e.g. a pre-convolved image, or a mask plane.
528 Returns
529 -------
530 results : `lsst.pipe.base.Struct`
531 The `~lsst.pipe.base.Struct` contains:
533 ``middle``
534 Convolved image, without the edges. (`lsst.afw.image.MaskedImage`)
535 ``sigma``
536 Gaussian sigma used for the convolution. (`float`)
537 """
538 self.metadata["doSmooth"] = doSmooth
539 sigma = psf.computeShape(psf.getAveragePosition()).getDeterminantRadius()
540 self.metadata["sigma"] = sigma
542 if not doSmooth:
543 middle = maskedImage.Factory(maskedImage, deep=True)
544 return pipeBase.Struct(middle=middle, sigma=sigma)
546 # Smooth using a Gaussian (which is separable, hence fast) of width sigma
547 # Make a SingleGaussian (separable) kernel with the 'sigma'
548 kWidth = self.calculateKernelSize(sigma)
549 self.metadata["smoothingKernelWidth"] = kWidth
550 gaussFunc = afwMath.GaussianFunction1D(sigma)
551 gaussKernel = afwMath.SeparableKernel(kWidth, kWidth, gaussFunc, gaussFunc)
553 convolvedImage = maskedImage.Factory(maskedImage.getBBox())
555 afwMath.convolve(convolvedImage, maskedImage, gaussKernel, afwMath.ConvolutionControl())
557 # Only search psf-smoothed part of frame
558 goodBBox = gaussKernel.shrinkBBox(convolvedImage.getBBox())
559 middle = convolvedImage.Factory(convolvedImage, goodBBox, afwImage.PARENT, False)
561 # Mark the parts of the image outside goodBBox as EDGE
562 self.setEdgeBits(maskedImage, goodBBox, maskedImage.getMask().getPlaneBitMask("EDGE"))
564 return pipeBase.Struct(middle=middle, sigma=sigma)
566 def applyThreshold(self, middle, bbox, factor=1.0, factorNeg=None):
567 r"""Apply thresholds to the convolved image
569 Identifies `~lsst.afw.detection.Footprint`\s, both positive and negative.
570 The threshold can be modified by the provided multiplication
571 ``factor``.
573 Parameters
574 ----------
575 middle : `lsst.afw.image.MaskedImage`
576 Convolved image to threshold.
577 bbox : `lsst.geom.Box2I`
578 Bounding box of unconvolved image.
579 factor : `float`, optional
580 Multiplier for the configured threshold.
581 factorNeg : `float` or `None`, optional
582 Multiplier for the configured threshold for negative detection
583 polarity. If `None`, will be set equal to ``factor`` (i.e. equal
584 to the factor used for positive detection polarity).
586 Returns
587 -------
588 results : `lsst.pipe.base.Struct`
589 The `~lsst.pipe.base.Struct` contains:
591 ``positive``
592 Positive detection footprints, if configured.
593 (`lsst.afw.detection.FootprintSet` or `None`)
594 ``negative``
595 Negative detection footprints, if configured.
596 (`lsst.afw.detection.FootprintSet` or `None`)
597 ``factor``
598 Multiplier for the configured threshold for positive detection
599 polarity. (`float`)
600 ``factorNeg``
601 Multiplier for the configured threshold for negative detection
602 polarity. (`float`)
603 """
604 if factorNeg is None:
605 factorNeg = factor
606 self.log.info("Threshold scaling factor for positive detections is: %.3f. For negative "
607 "detections it is: %.3f", factor, factorNeg)
608 results = pipeBase.Struct(positive=None, negative=None, factor=factor, factorNeg=factorNeg,
609 positiveThreshold=None, negativeThreshold=None)
610 # Detect the Footprints (peaks may be replaced if doTempLocalBackground)
611 if self.config.reEstimateBackground or self.config.thresholdPolarity != "negative":
612 results.positiveThreshold = self.makeThreshold(middle, "positive", factor=factor)
613 results.positive = afwDet.FootprintSet(
614 middle,
615 results.positiveThreshold,
616 "DETECTED",
617 self.config.minPixels
618 )
619 results.positive.setRegion(bbox)
620 if self.config.reEstimateBackground or self.config.thresholdPolarity != "positive":
621 results.negativeThreshold = self.makeThreshold(middle, "negative", factor=factorNeg)
622 results.negative = afwDet.FootprintSet(
623 middle,
624 results.negativeThreshold,
625 "DETECTED_NEGATIVE",
626 self.config.minPixels
627 )
628 results.negative.setRegion(bbox)
630 return results
632 def finalizeFootprints(self, mask, results, sigma, factor=1.0, factorNeg=None, growOverride=None):
633 """Finalize the detected footprints.
635 Grow the footprints, set the ``DETECTED`` and ``DETECTED_NEGATIVE``
636 mask planes, and log the results.
638 ``numPos`` (number of positive footprints), ``numPosPeaks`` (number
639 of positive peaks), ``numNeg`` (number of negative footprints),
640 ``numNegPeaks`` (number of negative peaks) entries are added to the
641 ``results`` struct.
643 Parameters
644 ----------
645 mask : `lsst.afw.image.Mask`
646 Mask image on which to flag detected pixels.
647 results : `lsst.pipe.base.Struct`
648 Struct of detection results, including ``positive`` and
649 ``negative`` entries; modified.
650 sigma : `float`
651 Gaussian sigma of PSF.
652 factor : `float`, optional
653 Multiplier for the configured threshold. Note that this is only
654 used here for logging purposes.
655 factorNeg : `float` or `None`, optional
656 Multiplier used for the negative detection polarity threshold.
657 If `None`, a factor equal to ``factor`` (i.e. equal to the one used
658 for positive detection polarity) is assumed. Note that this is only
659 used here for logging purposes.
660 growOverride : `float` or `None`, optional
661 Override to use for ``nSigmaToGrow``, regardless of the value set
662 in ``config.nSigmaToGrow``.
663 """
664 if growOverride is not None:
665 self.log.warning("config.nSigmaToGrow is set to %.2f, but the caller has set "
666 "growOverride to %.2f, so the footprints will be grown by "
667 "%.2f sigma.", self.config.nSigmaToGrow, growOverride, growOverride)
668 nSigmaToGrow = growOverride
669 else:
670 nSigmaToGrow = self.config.nSigmaToGrow
671 factorNeg = factor if factorNeg is None else factorNeg
672 for polarity, maskName in (("positive", "DETECTED"), ("negative", "DETECTED_NEGATIVE")):
673 fpSet = getattr(results, polarity)
674 if fpSet is None:
675 continue
676 if nSigmaToGrow > 0:
677 nGrow = int((self.config.nSigmaToGrow * sigma) + 0.5)
678 self.metadata["nGrow"] = nGrow
679 if self.config.combinedGrow:
680 fpSet = afwDet.FootprintSet(fpSet, nGrow, self.config.isotropicGrow)
681 else:
682 stencil = (afwGeom.Stencil.CIRCLE if self.config.isotropicGrow else
683 afwGeom.Stencil.MANHATTAN)
684 for fp in fpSet:
685 fp.dilate(nGrow, stencil)
686 fpSet.setMask(mask, maskName)
687 if not self.config.returnOriginalFootprints:
688 setattr(results, polarity, fpSet)
690 results.numPos = 0
691 results.numPosPeaks = 0
692 results.numNeg = 0
693 results.numNegPeaks = 0
694 positive = ""
695 negative = ""
697 if results.positive is not None:
698 results.numPos = len(results.positive.getFootprints())
699 results.numPosPeaks = sum(len(fp.getPeaks()) for fp in results.positive.getFootprints())
700 positive = " %d positive peaks in %d footprints" % (results.numPosPeaks, results.numPos)
701 if results.negative is not None:
702 results.numNeg = len(results.negative.getFootprints())
703 results.numNegPeaks = sum(len(fp.getPeaks()) for fp in results.negative.getFootprints())
704 negative = " %d negative peaks in %d footprints" % (results.numNegPeaks, results.numNeg)
706 self.log.info("Detected%s%s%s to %g +ve and %g -ve %s",
707 positive, " and" if positive and negative else "", negative,
708 self.config.thresholdValue*self.config.includeThresholdMultiplier*factor,
709 self.config.thresholdValue*self.config.includeThresholdMultiplier*factorNeg,
710 "DN" if self.config.thresholdType == "value" else "sigma")
712 def reEstimateBackground(self, maskedImage, backgrounds, backgroundToPhotometricRatio=None):
713 """Estimate the background after detection
715 Parameters
716 ----------
717 maskedImage : `lsst.afw.image.MaskedImage`
718 Image on which to estimate the background.
719 backgrounds : `lsst.afw.math.BackgroundList`
720 List of backgrounds; modified.
721 backgroundToPhotometricRatio : `lsst.afw.image.Image`, optional
722 Image to multiply a photometrically-flattened image by to obtain a
723 background-flattened image.
724 Only used if ``config.doApplyFlatBackgroundRatio`` is ``True``.
726 Returns
727 -------
728 bg : `lsst.afw.math.backgroundMI`
729 Empirical background model.
730 """
731 with backgroundFlatContext(
732 maskedImage,
733 self.config.doApplyFlatBackgroundRatio,
734 backgroundToPhotometricRatio=backgroundToPhotometricRatio,
735 ):
736 bg = self.background.fitBackground(maskedImage)
737 if self.config.adjustBackground:
738 self.log.warning("Fiddling the background by %g", self.config.adjustBackground)
739 bg += self.config.adjustBackground
740 self.log.info("Resubtracting the background after object detection (median background "
741 "value = %.2f)", np.median(bg.getImageF().array))
742 maskedImage -= bg.getImageF(self.background.config.algorithm,
743 self.background.config.undersampleStyle)
745 actrl = bg.getBackgroundControl().getApproximateControl()
746 backgrounds.append((bg, getattr(afwMath.Interpolate, self.background.config.algorithm),
747 bg.getAsUsedUndersampleStyle(), actrl.getStyle(), actrl.getOrderX(),
748 actrl.getOrderY(), actrl.getWeighting()))
749 return bg
751 def clearUnwantedResults(self, mask, results):
752 """Clear unwanted results from the Struct of results
754 If we specifically want only positive or only negative detections,
755 drop the ones we don't want, and its associated mask plane.
757 Parameters
758 ----------
759 mask : `lsst.afw.image.Mask`
760 Mask image.
761 results : `lsst.pipe.base.Struct`
762 Detection results, with ``positive`` and ``negative`` elements;
763 modified.
764 """
765 if self.config.thresholdPolarity == "positive":
766 if self.config.reEstimateBackground:
767 mask &= ~mask.getPlaneBitMask("DETECTED_NEGATIVE")
768 results.negative = None
769 elif self.config.thresholdPolarity == "negative":
770 if self.config.reEstimateBackground:
771 mask &= ~mask.getPlaneBitMask("DETECTED")
772 results.positive = None
774 @timeMethod
775 def detectFootprints(self, exposure, doSmooth=True, sigma=None, clearMask=True, expId=None,
776 background=None, backgroundToPhotometricRatio=None, factor=1.0, factorNeg=None):
777 """Detect footprints on an exposure.
779 Parameters
780 ----------
781 exposure : `lsst.afw.image.Exposure`
782 Exposure to process; DETECTED{,_NEGATIVE} mask plane will be
783 set in-place.
784 doSmooth : `bool`, optional
785 If True, smooth the image before detection using a Gaussian
786 of width ``sigma``, or the measured PSF width of ``exposure``.
787 Set to False when running on e.g. a pre-convolved image, or a mask
788 plane.
789 sigma : `float`, optional
790 Gaussian Sigma of PSF (pixels); used for smoothing and to grow
791 detections; if `None` then measure the sigma of the PSF of the
792 ``exposure``.
793 clearMask : `bool`, optional
794 Clear both DETECTED and DETECTED_NEGATIVE planes before running
795 detection.
796 expId : `dict`, optional
797 Exposure identifier; unused by this implementation, but used for
798 RNG seed by subclasses.
799 background : `lsst.afw.math.BackgroundList`, optional
800 Background that was already subtracted from the exposure; will be
801 modified in-place if ``reEstimateBackground=True``.
802 backgroundToPhotometricRatio : `lsst.afw.image.Image`, optional
803 Image to convert photometric-flattened image to
804 background-flattened image if ``reEstimateBackground=True`` and
805 exposure has been photometric-flattened.
806 factor : `float`, optional
807 Multiplier for the configured threshold for positive detection
808 polarity.
809 factorNeg : `float` or `None`, optional
810 Multiplier for the configured threshold for negative detection
811 polarity. If `None`, will be set equal to ``factor`` (i.e. equal
812 to the factor used for positive detection polarity).
814 Returns
815 -------
816 results : `lsst.pipe.base.Struct`
817 A `~lsst.pipe.base.Struct` containing:
819 ``positive``
820 Positive polarity footprints.
821 (`lsst.afw.detection.FootprintSet` or `None`)
822 ``negative``
823 Negative polarity footprints.
824 (`lsst.afw.detection.FootprintSet` or `None`)
825 ``numPos``
826 Number of footprints in positive or 0 if detection polarity was
827 negative. (`int`)
828 ``numNeg``
829 Number of footprints in negative or 0 if detection polarity was
830 positive. (`int`)
831 ``background``
832 Re-estimated background. `None` or the input ``background``
833 if ``reEstimateBackground==False``.
834 (`lsst.afw.math.BackgroundList`)
835 ``factor``
836 Multiplication factor applied to the configured threshold
837 for positive detection polarity. (`float`)
838 ``factorNeg``
839 Multiplication factor applied to the configured threshold
840 for negative detection polarity. (`float`)
841 """
842 maskedImage = exposure.maskedImage
844 if clearMask:
845 self.clearMask(maskedImage.getMask())
847 psf = self.getPsf(exposure, sigma=sigma)
848 with self.tempWideBackgroundContext(exposure):
849 convolveResults = self.convolveImage(maskedImage, psf, doSmooth=doSmooth)
850 middle = convolveResults.middle
851 sigma = convolveResults.sigma
852 self.removeBadPixels(middle)
854 results = self.applyThreshold(middle, maskedImage.getBBox(), factor=factor, factorNeg=factorNeg)
855 results.background = background if background is not None else afwMath.BackgroundList()
857 if self.config.doTempLocalBackground:
858 self.applyTempLocalBackground(exposure, middle, results)
859 self.finalizeFootprints(maskedImage.mask, results, sigma, factor=factor, factorNeg=factorNeg)
861 # Compute the significance of peaks after the peaks have been
862 # finalized and after local background correction/updatePeaks, so
863 # that the significance represents the "final" detection S/N.
864 results.positive = self.setPeakSignificance(middle, results.positive, results.positiveThreshold)
865 results.negative = self.setPeakSignificance(middle, results.negative, results.negativeThreshold,
866 negative=True)
868 if self.config.reEstimateBackground:
869 self.reEstimateBackground(
870 maskedImage,
871 results.background,
872 backgroundToPhotometricRatio=backgroundToPhotometricRatio,
873 )
875 self.clearUnwantedResults(maskedImage.getMask(), results)
877 self.display(exposure, results, middle)
879 return results
881 def removeBadPixels(self, middle):
882 """Set the significance of flagged pixels to zero.
884 Parameters
885 ----------
886 middle : `lsst.afw.image.Exposure`
887 Score or maximum likelihood difference image.
888 The image plane will be modified in place.
889 """
890 badPixelMask = middle.mask.getPlaneBitMask(self.config.excludeMaskPlanes)
891 badPixels = middle.mask.array & badPixelMask > 0
892 middle.image.array[badPixels] = 0
894 def setPeakSignificance(self, exposure, footprints, threshold, negative=False):
895 """Set the significance of each detected peak to the pixel value divided
896 by the appropriate standard-deviation for ``config.thresholdType``.
898 Only sets significance for "stdev" and "pixel_stdev" thresholdTypes;
899 we leave it undefined for "value" and "variance" as it does not have a
900 well-defined meaning in those cases.
902 Parameters
903 ----------
904 exposure : `lsst.afw.image.Exposure`
905 Exposure that footprints were detected on, likely the convolved,
906 local background-subtracted image.
907 footprints : `lsst.afw.detection.FootprintSet`
908 Footprints detected on the image.
909 threshold : `lsst.afw.detection.Threshold`
910 Threshold used to find footprints.
911 negative : `bool`, optional
912 Are we calculating for negative sources?
913 """
914 if footprints is None or footprints.getFootprints() == []:
915 return footprints
916 polarity = -1 if negative else 1
918 # All incoming footprints have the same schema.
919 mapper = afwTable.SchemaMapper(footprints.getFootprints()[0].peaks.schema)
920 mapper.addMinimalSchema(footprints.getFootprints()[0].peaks.schema)
921 mapper.addOutputField("significance", type=float,
922 doc="Ratio of peak value to configured standard deviation.")
924 # Copy the old peaks to the new ones with a significance field.
925 # Do this independent of the threshold type, so we always have a
926 # significance field.
927 newFootprints = afwDet.FootprintSet(footprints)
928 for old, new in zip(footprints.getFootprints(), newFootprints.getFootprints()):
929 newPeaks = afwDet.PeakCatalog(mapper.getOutputSchema())
930 newPeaks.extend(old.peaks, mapper=mapper)
931 new.getPeaks().clear()
932 new.setPeakCatalog(newPeaks)
934 # Compute the significance values.
935 if self.config.thresholdType == "pixel_stdev":
936 for footprint in newFootprints.getFootprints():
937 footprint.updatePeakSignificance(exposure.variance, polarity)
938 elif self.config.thresholdType == "stdev":
939 sigma = threshold.getValue() / self.config.thresholdValue
940 for footprint in newFootprints.getFootprints():
941 footprint.updatePeakSignificance(polarity*sigma)
942 else:
943 for footprint in newFootprints.getFootprints():
944 for peak in footprint.peaks:
945 peak["significance"] = 0
947 return newFootprints
949 def makeThreshold(self, image, thresholdParity, factor=1.0):
950 """Make an afw.detection.Threshold object corresponding to the task's
951 configuration and the statistics of the given image.
953 Parameters
954 ----------
955 image : `afw.image.MaskedImage`
956 Image to measure noise statistics from if needed.
957 thresholdParity: `str`
958 One of "positive" or "negative", to set the kind of fluctuations
959 the Threshold will detect.
960 factor : `float`
961 Factor by which to multiply the configured detection threshold.
962 This is useful for tweaking the detection threshold slightly.
964 Returns
965 -------
966 threshold : `lsst.afw.detection.Threshold`
967 Detection threshold.
968 """
969 parity = False if thresholdParity == "negative" else True
970 thresholdValue = self.config.thresholdValue
971 thresholdType = self.config.thresholdType
972 if self.config.thresholdType == 'stdev':
973 bad = image.getMask().getPlaneBitMask(self.config.statsMask)
974 sctrl = afwMath.StatisticsControl()
975 sctrl.setAndMask(bad)
976 stats = afwMath.makeStatistics(image, afwMath.STDEVCLIP, sctrl)
977 thresholdValue *= stats.getValue(afwMath.STDEVCLIP)
978 thresholdType = 'value'
980 threshold = afwDet.createThreshold(thresholdValue*factor, thresholdType, parity)
981 threshold.setIncludeMultiplier(self.config.includeThresholdMultiplier)
982 self.log.debug("Detection threshold: %s", threshold)
983 return threshold
985 def updatePeaks(self, fpSet, image, threshold):
986 """Update the Peaks in a FootprintSet by detecting new Footprints and
987 Peaks in an image and using the new Peaks instead of the old ones.
989 Parameters
990 ----------
991 fpSet : `afw.detection.FootprintSet`
992 Set of Footprints whose Peaks should be updated.
993 image : `afw.image.MaskedImage`
994 Image to detect new Footprints and Peak in.
995 threshold : `afw.detection.Threshold`
996 Threshold object for detection.
998 Input Footprints with fewer Peaks than self.config.nPeaksMaxSimple
999 are not modified, and if no new Peaks are detected in an input
1000 Footprint, the brightest original Peak in that Footprint is kept.
1001 """
1002 for footprint in fpSet.getFootprints():
1003 oldPeaks = footprint.getPeaks()
1004 if len(oldPeaks) <= self.config.nPeaksMaxSimple:
1005 continue
1006 # We detect a new FootprintSet within each non-simple Footprint's
1007 # bbox to avoid a big O(N^2) comparison between the two sets of
1008 # Footprints.
1009 sub = image.Factory(image, footprint.getBBox())
1010 fpSetForPeaks = afwDet.FootprintSet(
1011 sub,
1012 threshold,
1013 "", # don't set a mask plane
1014 self.config.minPixels
1015 )
1016 newPeaks = afwDet.PeakCatalog(oldPeaks.getTable())
1017 for fpForPeaks in fpSetForPeaks.getFootprints():
1018 for peak in fpForPeaks.getPeaks():
1019 if footprint.contains(peak.getI()):
1020 newPeaks.append(peak)
1021 if len(newPeaks) > 0:
1022 del oldPeaks[:]
1023 oldPeaks.extend(newPeaks)
1024 else:
1025 del oldPeaks[1:]
1027 @staticmethod
1028 def setEdgeBits(maskedImage, goodBBox, edgeBitmask):
1029 """Set the edgeBitmask bits for all of maskedImage outside goodBBox
1031 Parameters
1032 ----------
1033 maskedImage : `lsst.afw.image.MaskedImage`
1034 Image on which to set edge bits in the mask.
1035 goodBBox : `lsst.geom.Box2I`
1036 Bounding box of good pixels, in ``LOCAL`` coordinates.
1037 edgeBitmask : `lsst.afw.image.MaskPixel`
1038 Bit mask to OR with the existing mask bits in the region
1039 outside ``goodBBox``.
1040 """
1041 msk = maskedImage.getMask()
1043 mx0, my0 = maskedImage.getXY0()
1044 for x0, y0, w, h in ([0, 0,
1045 msk.getWidth(), goodBBox.getBeginY() - my0],
1046 [0, goodBBox.getEndY() - my0, msk.getWidth(),
1047 maskedImage.getHeight() - (goodBBox.getEndY() - my0)],
1048 [0, 0,
1049 goodBBox.getBeginX() - mx0, msk.getHeight()],
1050 [goodBBox.getEndX() - mx0, 0,
1051 maskedImage.getWidth() - (goodBBox.getEndX() - mx0), msk.getHeight()],
1052 ):
1053 edgeMask = msk.Factory(msk, lsst.geom.BoxI(lsst.geom.PointI(x0, y0),
1054 lsst.geom.ExtentI(w, h)), afwImage.LOCAL)
1055 edgeMask |= edgeBitmask
1057 @contextmanager
1058 def tempWideBackgroundContext(self, exposure):
1059 """Context manager for removing wide (large-scale) background
1061 Removing a wide (large-scale) background helps to suppress the
1062 detection of large footprints that may overwhelm the deblender.
1063 It does, however, set a limit on the maximum scale of objects.
1065 The background that we remove will be restored upon exit from
1066 the context manager.
1068 Parameters
1069 ----------
1070 exposure : `lsst.afw.image.Exposure`
1071 Exposure on which to remove large-scale background.
1073 Returns
1074 -------
1075 context : context manager
1076 Context manager that will ensure the temporary wide background
1077 is restored.
1078 """
1079 doTempWideBackground = self.config.doTempWideBackground
1080 if doTempWideBackground:
1081 self.log.info("Applying temporary wide background subtraction")
1082 original = exposure.maskedImage.image.array[:].copy()
1083 self.tempWideBackground.run(exposure).background
1084 # Remove NO_DATA regions (e.g., edge of the field-of-view); these can cause detections after
1085 # subtraction because of extrapolation of the background model into areas with no constraints.
1086 image = exposure.maskedImage.image
1087 mask = exposure.maskedImage.mask
1088 noData = mask.array & mask.getPlaneBitMask("NO_DATA") > 0
1089 isGood = mask.array & mask.getPlaneBitMask(self.config.statsMask) == 0
1090 image.array[noData] = np.median(image.array[~noData & isGood])
1091 try:
1092 yield
1093 finally:
1094 if doTempWideBackground:
1095 exposure.maskedImage.image.array[:] = original
1098def addExposures(exposureList):
1099 """Add a set of exposures together.
1101 Parameters
1102 ----------
1103 exposureList : `list` of `lsst.afw.image.Exposure`
1104 Sequence of exposures to add.
1106 Returns
1107 -------
1108 addedExposure : `lsst.afw.image.Exposure`
1109 An exposure of the same size as each exposure in ``exposureList``,
1110 with the metadata from ``exposureList[0]`` and a masked image equal
1111 to the sum of all the exposure's masked images.
1112 """
1113 exposure0 = exposureList[0]
1114 image0 = exposure0.getMaskedImage()
1116 addedImage = image0.Factory(image0, True)
1117 addedImage.setXY0(image0.getXY0())
1119 for exposure in exposureList[1:]:
1120 image = exposure.getMaskedImage()
1121 addedImage += image
1123 addedExposure = exposure0.Factory(addedImage, exposure0.getWcs())
1124 return addedExposure