def mask_catalog_columns(catalog, idx, columns=None,
catname=None, new_columns=None,
catalog_kwargs=None):
"""
Mask all of the rows in a table (or subset of columns in a table)
with a masked index and (optionally) rename the columns.
Parameters
----------
catalog: `~astropy.table.Table` or `~Catalog`
Catalog or Table to be masked
idx: `~numpy.ma.array`
Masked array of indices to use for updating the catalog
columns: list of strings, optional
Columns to include in the masked table. If ``columns is None``
(default) then all of the columns are included in the masked catalog.
catname: str, optional
Name of catalog. This is only necessary if you with to rename the
columns of the catalog for stacking later. See ``new_columns``
for more.
new_columns: list of strings, optional
New names for the columns. This may be useful if you are combining
catalogs and want to standardize the column names. If
``new_columns is None`` (default) then if a ``catname`` is provided all
of the columns are renamed to 'columnname_catname', otherwise
the original column names are used.
catalog_kwargs: dict
If the result is desired to be a `~astropyp.catalog.Catalog` then
these are the kwargs to specify when initializing the catalog
(for example names of the ra,dec,x,y columns). Otherwise
an `~astropy.table.Table` is returned
Returns
-------
tbl: `~astropy.table.Table` or `~astropyp.catalog.Catalog`
Catalog created by applying the masked index. The type of object
returned depends on if ``catalog_kwargs is None`` (default), which
returns a Table. Otherwise a Catalog is returned.
"""
from astropy.table import Table
if isinstance(catalog, Catalog):
tbl = catalog.sources
else:
tbl = catalog
new_tbl = Table(masked=True)
if columns is None:
columns = tbl.columns.keys()
if new_columns is None:
if catname is None:
new_columns = columns
else:
new_columns = ['{0}_{1}'.format(col,catname) for col in columns]
for n, col in enumerate(columns):
new_tbl[new_columns[n]] = misc.update_ma_idx(tbl[col], idx)
if catalog_kwargs is not None:
new_tbl = Catalog(new_tbl, **catalog_kwargs)
return new_tbl
def merge_catalogs(self, good_indices=None, save_catalog=True):
"""
Combine catalogs from each image into a master catalog that
matches all of the source (by default, although specifying
``good_indices`` allows to only merge a subset of each
catalog)
Parameters
----------
good_indices: string or list of arrays, optional
If good_indices is a string, index ``ccd.indices[good_indices]``
will be used for each CCD to filter the sources. Otherwise
good_indices should be a list of indices to use for each CCD.
If no god_indices are specified then all sources will be
merged
save_catalog: bool
Whether or not to save the catalog
Result
------
catalog: `~astropyp.catalog.Catalog`
Catalog of merged positions and indices for the sources
in each image.
"""
import astropyp.catalog
from astropyp.utils.misc import update_ma_idx
from astropy import table
from astropy.extern import six
if good_indices is None:
all_ra = [ccd.catalog.ra for ccd in self.ccds]
all_dec = [ccd.catalog.dec for ccd in self.ccds]
else:
if isinstance(good_indices, six.string_types):
good_indices = [ccd.indices[good_indices] for ccd in self.ccds]
all_ra = [ccd.catalog.ra[good_indices[n]]
for n, ccd in enumerate(self.ccds)]
all_dec = [ccd.catalog.dec[good_indices[n]]
for n, ccd in enumerate(self.ccds)]
indices, matched = astropyp.catalog.get_merged_indices(all_ra, all_dec)
all_x = []
all_y = []
for n,ccd in enumerate(self.ccds):
ccd.indices['merge'] = indices[n]
if n!=self.ref_index:
x0,y0 = self.tx_solutions[
(n,self.ref_index)].transform_coords(ccd.catalog)
else:
x0,y0 = (ccd.catalog.x, ccd.catalog.y)
if good_indices is not None:
x0 = x0[good_indices[n]]
y0 = y0[good_indices[n]]
all_x.append(x0)
all_y.append(y0)
all_x = [update_ma_idx(all_x[n], self.ccds[n].indices['merge'])
for n in self.ccd_indices]
all_y = [update_ma_idx(all_y[n], self.ccds[n].indices['merge'])
for n in self.ccd_indices]
sources = table.Table(masked=True)
sources['x'],sources['y'] = astropyp.catalog.combine_coordinates(
all_x, all_y, 'mean')
for n,ccd in enumerate(self.ccds):
x_col = 'x_{0}'.format(n)
y_col = 'y_{0}'.format(n)
sources[x_col] = all_x[n]
sources[y_col] = all_y[n]
if good_indices is None:
src_idx = self.ccds[n].catalog['src_idx']
else:
src_idx = self.ccds[n].catalog['src_idx'][good_indices[n]]
src_idx = update_ma_idx(src_idx,
self.ccds[n].indices['merge'])
sources['idx_{0}'.format(n)] = src_idx
x_diff = sources[x_col]-sources['x']
y_diff = sources[y_col]-sources['y']
logger.info('x rms: {0}'.format(
np.sqrt(np.sum(x_diff)**2/len(sources))))
logger.info('y rms: {0}'.format(
np.sqrt(np.sum(y_diff)**2/len(sources))))
catalog = astropyp.catalog.Catalog(sources)
if save_catalog:
self.merged_catalog = catalog
return catalog
def perform_psf_photometry(self, separation=None,
verbose=False, fit_position=True, pos_range=0, indices=None,
kd_tree=None, exptime=None, pool_size=None, stack_method='mean',
save_catalog=True, single_thread=False):
"""
Perform PSF photometry on all of the sources in the catalog,
or if indices is specified, a subset of sources.
Parameters
----------
separation: float, optional
Separation (in pixels) for members to be considered
part of the same group *Default=1.5*psf width*
verbose: bool, optional
Whether or not to show info about the fit progress.
*Default=False*
fit_position: bool, optional
Whether or not to fit the position along with the
amplitude of each source. *Default=True*
pos_range: int, optional
Maximum number of pixels (in image pixels) that
a sources position can be changed. If ``pos_range=0``
no bounds will be set. *Default=0*
indices: `~numpy.ndarray` or string, optional
Indices for sources to calculate PSF photometry.
It is often advantageous to remove sources with
bad pixels and sublinear flux to save processing time.
All sources not included in indices will have their
psf flux set to NaN. This can either be an array of
indices for the positions in self.catalog or
the name of a saved index in self.indices
"""
import astropyp.catalog
from astropyp.phot.psf import SinglePSF
# If no exposure time was passed to the stack and all of the
# ccds had the same exposure time, use that to calcualte the
# psf magnitude
if exptime is None:
if hasattr(self, 'exptime'):
exptime = self.exptime
else:
same_exptime = True
exptime = np.nan
for ccd in self.ccds:
if np.isnan(exptime):
exptime = ccd.exptime
elif ccd.exptime != exptime:
same_exptime = False
if not same_exptime:
exptime = None
if hasattr(self.catalog, 'peak'):
peak = self.catalog.peak
else:
from astropyp.utils.misc import update_ma_idx
peak = [
update_ma_idx(self.ccds[n].catalog['peak'],
self.catalog['idx_'+str(n)])
for n in self.ccd_indices]
peak = np.ma.mean(peak, axis=0)
# Get the positions and estimated amplitudes of
# the sources to fit
if indices is not None:
positions = zip(self.catalog.x[indices],
self.catalog.y[indices])
amplitudes = peak[indices]
else:
positions = zip(self.catalog.x, self.catalog.y)
amplitudes = peak
src_count = len(positions)
src_indices = np.arange(0,len(self.catalog.sources),1)
all_positions = np.array(zip(self.catalog.x, self.catalog.y))
all_amplitudes = peak
total_sources = len(all_amplitudes)
src_psfs = []
psf_flux = np.ones((total_sources,))*np.nan
psf_flux_err = np.ones((total_sources,))*np.nan
psf_x = np.ones((total_sources,))*np.nan
psf_y = np.ones((total_sources,))*np.nan
new_amplitudes = np.ones((total_sources,))*np.nan
# Find nearest neighbors to avoid contamination by nearby sources
if separation is not None:
if not hasattr(self, 'kdtree'):
from scipy import spatial
KDTree = spatial.cKDTree
self.kd_tree = KDTree(all_positions)
idx, nidx = astropyp.catalog.find_neighbors(separation,
kd_tree=self.kd_tree)
# Get parameters to pass to the pool manager
imgs = []
dqmasks = []
tx_solutions = []
for n,ccd in enumerate(self.ccds):
imgs.append(ccd.img)
dqmasks.append(ccd.dqmask)
if n!= self.ref_index:
tx_solutions.append(self.tx_solutions[(self.ref_index, n)])
else:
tx_solutions.append(None)
if pool_size is None:
pool_size = multiprocessing.cpu_count()
# Create a pool with the static (for all sources)
# variables to speed up processing
pool = multiprocessing.Pool(
processes=pool_size,
initializer=_init_multiprocess,
initargs=(imgs, dqmasks, tx_solutions, self.psf.shape,
self.ref_index, stack_method, self.subsampling,
self.psf))
pool_args = []
# Fit each source to the PSF, calcualte its flux, and its new
# position
for n in range(len(positions)):
if indices is not None:
src_idx = src_indices[indices][n]
else:
src_idx = src_indices[n]
if separation is not None:
n_indices = nidx[idx==src_idx]
neighbor_positions = all_positions[n_indices]
neighbor_amplitudes = all_amplitudes[n_indices]
else:
neighbor_positions = []
neighbor_amplitudes = []
pool_args.append((amplitudes[n],
(positions[n][1], positions[n][0]),
self.subsampling,
neighbor_positions,
neighbor_amplitudes))
result = pool.map(_stack_psf_worker, pool_args)
pool.close()
pool.join()
# Update the various psf array parameters
for n in range(len(positions)):
if indices is not None:
src_idx = src_indices[indices][n]
else:
src_idx = src_indices[n]
psf_flux[src_idx] = result[n][0]
psf_flux_err[src_idx] = result[n][1]
new_amplitudes[src_idx] = result[n][2]
psf_x[src_idx], psf_y[src_idx] = result[n][3]
# Save the psf derived quantities in the catalog
# Ignore divide by zero errors that occur when sources
# have zero psf flux (i.e. bad sources)
if save_catalog:
np_err = np.geterr()
np.seterr(divide='ignore')
self.catalog.sources['psf_flux'] = psf_flux
self.catalog.sources['psf_flux_err'] = psf_flux_err
if exptime is not None:
psf_mag = -2.5*np.log10(psf_flux/exptime)
self.catalog.sources['psf_mag'] = psf_mag
self.catalog.sources['psf_mag_err'] = psf_flux_err/psf_flux
self.catalog.sources['psf_x'] = psf_x
self.catalog.sources['psf_y'] = psf_y
np.seterr(**np_err)
return psf_flux, psf_flux_err
def get_all_merged_indices(all_coord1, all_coord2, pool_size=None,
separation=1/3600., merge_type='outer'):
"""
Get masked indices for a set of ra,dec that merge the
sources together using an outer join
Parameters
----------
all_coord1: list of array-like
List of arrays of values for the first coordinate (usually RA or X)
all_coord2: list of array-like
List of arrays of values for the second coordinate (usually DEC or Y)
pool_size: int, optional
Number of processors to use to match coordinates. If
``pool_size is None`` (default) then the maximum number
of processors is used.
separation: float, optional
Maximum distance between two coordinates for a match. The default
is ``1/3600``, or 1 arcsec.
merge_type: str, optional
Type of merge to use. This must be 'outer','inner', 'left', or 'right'.
The default is 'outer'.
Returns
-------
indices: list of masked arrays
Indices to match each catalog in all_coord1, all_coord2 to the
master catalog
matched: array
Indices of rows that has an entry for *every* set of coordinates
all_duplicates: array
Indices of rows that have duplicate values
mean_coord1: array
Average coord1 for each row
mean_coord2: array
Average coord2 for each row
"""
from astropyp.utils import misc
if merge_type not in ['outer','inner','left','right']:
raise ValueError(
"merge_type must be 'outer','inner', 'left', or 'right'")
# Initialize indices and coordinates
indices = [np.ma.array([n for n in
range(all_coord1[m].shape[0])], dtype=int)
for m in range(len(all_coord1))]
mean_coord1 = np.ma.array(all_coord1[0])
mean_coord2 = np.ma.array(all_coord2[0])
all_duplicates = np.zeros(mean_coord1.shape, dtype=bool)
# Create merged indices
for n in range(1,len(all_coord1)):
idx0, idx1, duplicates = get_merged_indices(
(all_coord1[n],all_coord2[n]),
ref_coords=(mean_coord1,mean_coord2),
pool_size=pool_size, separation=separation)
new_idx, new_unmatched = idx0
ref_idx, ref_unmatched = idx1
# Update list of duplicates
duplicates_unmatched = all_duplicates[ref_unmatched]
all_duplicates = all_duplicates[ref_idx]
all_duplicates[duplicates] = True
# Update indices
if merge_type=='outer' or merge_type=='left':
ref_idx = np.hstack([ref_idx, ref_unmatched])
new_idx = np.hstack([new_idx,
-np.ones(ref_unmatched.shape, dtype=int)])
all_duplicates = np.hstack([all_duplicates, duplicates_unmatched])
if merge_type=='outer' or merge_type=='right':
ref_idx = np.hstack([ref_idx,
-np.ones(new_unmatched.shape, dtype=int)])
new_idx = np.hstack([new_idx, new_unmatched])
all_duplicates = np.hstack([all_duplicates,
np.zeros(new_unmatched.shape, dtype=bool)])
# Mask indices
ref_mask = ref_idx<0
new_mask = new_idx<0
ref_idx = np.ma.array(ref_idx, mask=ref_mask)
new_idx = np.ma.array(new_idx, mask=new_mask)
# Update the mean coordinate values
mean_coord1 = misc.update_ma_idx(mean_coord1,ref_idx)
mean_coord2 = misc.update_ma_idx(mean_coord2,ref_idx)
new_coord1 = misc.update_ma_idx(all_coord1[n],new_idx)
new_coord2 = misc.update_ma_idx(all_coord2[n],new_idx)
mean_coord1 = np.ma.mean(
np.ma.vstack([mean_coord1, new_coord1]), axis=0)
mean_coord2 = np.ma.mean(
np.ma.vstack([mean_coord2, new_coord2]), axis=0)
# Update all of the indices with the new matches
for m in range(n):
indices[m] = misc.update_ma_idx(indices[m],ref_idx)
indices[n] = new_idx
matched = np.sum([i.mask for i in indices],axis=0)==0
return indices, matched, all_duplicates, mean_coord1, mean_coord2
