def do_photometry(params,
extname='newflux',
star_file='star_positions',
psf_file='psf.fits',
star_positions=None,
reference_image='ref.fits'):
#
# Determine our list of files
#
all_files = os.listdir(params.loc_data)
all_files.sort()
files = []
for f in all_files:
if fnmatch.fnmatch(f, params.name_pattern):
g = DS.Observation(params.loc_data + os.path.sep + f, params)
if g.fw > 0.0:
files.append(g)
ref = DS.Observation(params.loc_output + os.path.sep + reference_image,
params)
ref.register(ref, params)
#
# Detect stars and compute the PSF if necessary
#
psf_file = params.loc_output + os.path.sep + psf_file
star_file = params.loc_output + os.path.sep + star_file
print psf_file
print os.path.exists(psf_file)
print star_file
print os.path.exists(star_file)
if not (os.path.exists(psf_file)) or not (os.path.exists(star_file)):
stars = PH.compute_psf_image(params, ref, psf_image=psf_file)
if star_positions is None:
if os.path.exists(star_file):
star_positions = np.genfromtxt(star_file)[:, :2]
else:
star_positions = stars[:, 0:2]
#
# Group the stars by location
#
star_group_boundaries = None
detector_mean_positions_x = None
detector_mean_positions_y = None
star_sort_index,star_group_boundaries,detector_mean_positions_x,detector_mean_positions_y = \
PH.group_stars_ccd(params,star_positions,params.loc_output+os.path.sep+reference_image)
star_positions = star_positions[star_sort_index]
star_unsort_index = np.argsort(star_sort_index)
#
# Process the reference image
#
print 'Processing', reference_image
ref = DS.Observation(params.loc_output + os.path.sep + reference_image,
params)
#reg = Observation(params.loc_data+os.path.sep+
# params.registration_image,params)
mask, _ = IO.read_fits_file(params.loc_output + os.path.sep + 'mask_' +
reference_image)
variance, _ = IO.read_fits_file(params.loc_output + os.path.sep + 'var_' +
reference_image)
ref.mask = mask
ref.inv_variance = 1.0 / variance + (1 - mask)
ref.register(ref, params)
smask = IM.compute_saturated_pixel_mask(ref.image, params)
ref.inv_variance += (1 - (smask * mask)) * 1.e-12
ktable = params.loc_output + os.path.sep + 'k_' + os.path.basename(
reference_image)
kernelIndex, extendedBasis, c, params = IO.read_kernel_table(
ktable, params)
kernelRadius = np.max(kernelIndex[:, 0]) + 1
if np.sum(extendedBasis) > 0:
kernelRadius += 1
print 'kernelIndex', kernelIndex
print 'extendedBasis', extendedBasis
print 'coeffs', c
print 'kernelRadius', kernelRadius
print 'star_positions', star_positions.shape
phot_target, _ = IO.read_fits_file(params.loc_output + os.path.sep +
'clean_' + reference_image)
ref.flux, ref.dflux = CIF.photom_all_stars_simultaneous(
phot_target, ref.inv_variance, star_positions, psf_file, c,
kernelIndex, extendedBasis, kernelRadius, params,
star_group_boundaries, detector_mean_positions_x,
detector_mean_positions_y)
if isinstance(ref.flux, np.ndarray):
if not (params.use_GPU):
print 'ungrouping fluxes'
ref.flux = ref.flux[star_unsort_index].copy()
ref.dflux = ref.dflux[star_unsort_index].copy()
print ref.flux.shape, star_positions.shape
np.savetxt(
params.loc_output + os.path.sep + reference_image + '.' + extname,
np.vstack((ref.flux, ref.dflux)))
#
# Process difference images
#
for f in files:
if not (os.path.exists(params.loc_output + os.path.sep + f.name + '.' +
extname)):
print 'Processing', f.name
target = f.name
dtarget = params.loc_output + os.path.sep + 'd_' + os.path.basename(
target)
ntarget = params.loc_output + os.path.sep + 'n_' + os.path.basename(
target)
ztarget = params.loc_output + os.path.sep + 'z_' + os.path.basename(
target)
ktable = params.loc_output + os.path.sep + 'k_' + os.path.basename(
target)
if os.path.exists(dtarget) and os.path.exists(
ntarget) and os.path.exists(ktable):
norm, h = IO.read_fits_file(ntarget)
diff, h = IO.read_fits_file(dtarget)
mask, h = IO.read_fits_file(ztarget)
inv_var = (norm / diff)**2 + (1 - mask)
kernelIndex, extendedBasis, c, params = IO.read_kernel_table(
ktable, params)
kernelRadius = np.max(kernelIndex[:, 0]) + 1
if np.sum(extendedBasis) > 0:
kernelRadius += 1
print 'kernelIndex', kernelIndex
print 'extendedBasis', extendedBasis
print 'coeffs', c
print 'kernelRadius', kernelRadius
IO.write_image(diff,
params.loc_output + os.path.sep + 'diff1.fits')
diff = IM.undo_photometric_scale(diff, c, params.pdeg)
IO.write_image(diff,
params.loc_output + os.path.sep + 'diff2.fits')
IO.write_image(
inv_var, params.loc_output + os.path.sep + 'inv_var.fits')
IO.write_kernel_table(
params.loc_output + os.path.sep + 'ktable.fits',
kernelIndex, extendedBasis, c, params)
flux, dflux = CI.photom_all_stars(
diff, inv_var, star_positions, psf_file, c, kernelIndex,
extendedBasis, kernelRadius, params, star_group_boundaries,
detector_mean_positions_x, detector_mean_positions_y)
print 'flux[100:110]:'
print flux[100:110]
if isinstance(flux, np.ndarray):
if not (params.use_GPU):
print 'ungrouping fluxes'
flux = flux[star_unsort_index].copy()
dflux = dflux[star_unsort_index].copy()
print 'unsort flux[100:110]:'
print flux[100:110]
np.savetxt(
params.loc_output + os.path.sep + f.name + '.' +
extname,
np.vstack((flux, dflux)).T)
sys.exit(0)
def do_photometry(params,
extname='newflux',
star_file='star_positions',
psf_file='psf.fits',
star_positions=None,
reference_image='ref.fits'):
#
# Determine our list of files
#
all_files = os.listdir(params.loc_data)
all_files.sort()
files = []
for f in all_files:
if fnmatch.fnmatch(f, params.name_pattern):
g = DS.Observation(params.loc_data + os.path.sep + f, params)
if g.fw > 0.0:
files.append(g)
ref = DS.Observation(params.loc_output + os.path.sep + reference_image,
params)
ref.register(ref, params)
#
# Detect stars and compute the PSF if necessary
#
if params.do_photometry:
psf_file = params.loc_output + os.path.sep + psf_file
if os.path.exists(params.star_file):
star_pos = np.genfromtxt(params.star_file)[:, 1:3]
if not (os.path.exists(psf_file)):
stars = PH.compute_psf_image(params, ref, psf_image=psf_file)
else:
if not (os.path.exists(star_file)):
stars = PH.compute_psf_image(params, ref, psf_image=psf_file)
star_pos = stars[:, 0:2]
np.savetxt(star_file, star_pos)
else:
star_pos = np.genfromtxt(star_file)
if not (os.path.exists(psf_file)):
stars = PH.compute_psf_image(params,
ref,
psf_image=psf_file)
#
# Have we been passed an array of star positions?
#
if star_positions == None:
star_positions = star_pos
#
# If we are using a CPU, group the stars by location
#
star_group_boundaries = None
detector_mean_positions_x = None
detector_mean_positions_y = None
if not (params.use_GPU):
star_sort_index, star_group_boundaries, detector_mean_positions_x, detector_mean_positions_y = PH.group_stars_ccd(
params, star_positions,
params.loc_output + os.path.sep + reference_image)
star_positions = star_positions[star_sort_index]
star_unsort_index = np.argsort(star_sort_index)
#
# Process the reference image
#
print('Processing', reference_image)
ref = DS.Observation(params.loc_output + os.path.sep + reference_image,
params)
# reg = Observation(params.loc_data+os.path.sep+
# params.registration_image,params)
ref.register(ref, params)
smask = IM.compute_saturated_pixel_mask(ref.image, 6, params)
ref.inv_variance += 1 - smask
ktable = params.loc_output + os.path.sep + 'k_' + os.path.basename(
reference_image)
kernelIndex, extendedBasis, c, params = IO.read_kernel_table(
ktable, params)
kernelRadius = np.max(kernelIndex[:, 0]) + 1
if np.sum(extendedBasis) > 0:
kernelRadius += 1
print('kernelIndex', kernelIndex)
print('extendedBasis', extendedBasis)
print('coeffs', c)
print('kernelRadius', kernelRadius)
phot_target = ref.image
ref.flux, ref.dflux = CI.photom_all_stars(
phot_target, ref.inv_variance, star_positions, psf_file, c,
kernelIndex, extendedBasis, kernelRadius, params,
star_group_boundaries, detector_mean_positions_x,
detector_mean_positions_y)
if isinstance(ref.flux, np.ndarray):
if not (params.use_GPU):
print('ungrouping fluxes')
ref.flux = ref.flux[star_unsort_index].copy()
ref.dflux = ref.dflux[star_unsort_index].copy()
np.savetxt(
params.loc_output + os.path.sep + reference_image + '.' + extname,
np.vstack((ref.flux, ref.dflux)).T)
#
# Process difference images
#
for f in files:
if not (os.path.exists(params.loc_output + os.path.sep + f.name + '.' +
extname)):
print('Processing', f.name)
target = f.name
dtarget = params.loc_output + os.path.sep + 'd_' + os.path.basename(
target)
ntarget = params.loc_output + os.path.sep + 'n_' + os.path.basename(
target)
ztarget = params.loc_output + os.path.sep + 'z_' + os.path.basename(
target)
ktable = params.loc_output + os.path.sep + 'k_' + os.path.basename(
target)
if os.path.exists(dtarget) and os.path.exists(
ntarget) and os.path.exists(ktable):
norm, h = IO.read_fits_file(ntarget)
diff, h = IO.read_fits_file(dtarget)
mask, h = IO.read_fits_file(ztarget)
inv_var = (norm / diff)**2 + (1 - mask)
kernelIndex, extendedBasis, c, params = IO.read_kernel_table(
ktable, params)
kernelRadius = np.max(kernelIndex[:, 0]) + 1
if np.sum(extendedBasis) > 0:
kernelRadius += 1
print('kernelIndex', kernelIndex)
print('extendedBasis', extendedBasis)
print('coeffs', c)
print('kernelRadius', kernelRadius)
diff = IM.undo_photometric_scale(diff, c, params.pdeg)
flux, dflux = PH.photom_all_stars(
diff, inv_var, star_positions, psf_file, c, kernelIndex,
extendedBasis, kernelRadius, params, star_group_boundaries,
detector_mean_positions_x, detector_mean_positions_y)
if isinstance(flux, np.ndarray):
if not (params.use_GPU):
print('ungrouping fluxes')
flux = flux[star_unsort_index].copy()
dflux = dflux[star_unsort_index].copy()
np.savetxt(
params.loc_output + os.path.sep + f.name + '.' +
extname,
np.vstack((flux, dflux)).T)
def photom_variable_star(x0,
y0,
params,
patch_half_width=15,
converge=True,
save_stamps=False,
stamp_prefix='mosaic',
locate=True,
locate_iterations=2,
locate_half_width=14,
q_sigma_threshold=1.0,
locate_date_range=None):
from astropy.io import fits
def save_mosaic(stack, nfiles, patch_size, name, diff_std, threshold):
stamps_per_row = int(np.sqrt(nfiles))
nrows = (nfiles - 1) / stamps_per_row + 1
mx = stamps_per_row * (patch_size + 1) + 1
my = nrows * (patch_size + 1) + 1
mosaic = np.ones((my, mx)) * 1000.0
for i in range(nfiles):
mosaic[(i/stamps_per_row)*(patch_size+1)+1:(i/stamps_per_row+1)*(patch_size+1), \
(i%stamps_per_row)*(patch_size+1)+1:(i%stamps_per_row+1)*(patch_size+1)] \
= stack[i,:,:]
if diff_std[i] > threshold:
mosaic[(i/stamps_per_row)*(patch_size+1)+1:(i/stamps_per_row+1)*(patch_size+1), \
(i%stamps_per_row)*(patch_size+1)+1] = -1000.0
mosaic[(i/stamps_per_row)*(patch_size+1)+1:(i/stamps_per_row+1)*(patch_size+1), \
(i%stamps_per_row+1)*(patch_size+1)-1] = -1000.0
mosaic[(i/stamps_per_row)*(patch_size+1)+1, \
(i%stamps_per_row)*(patch_size+1)+1:(i%stamps_per_row+1)*(patch_size+1)] = -1000.0
mosaic[(i/stamps_per_row+1)*(patch_size+1)-1, \
(i%stamps_per_row)*(patch_size+1)+1:(i%stamps_per_row+1)*(patch_size+1)] = -1000.0
IO.write_image(mosaic, name)
# Obtain a list of files
all_files = os.listdir(params.loc_data)
all_files.sort()
filenames = []
nfiles = 0
print 'Searching in', params.loc_output, 'for', params.name_pattern
for f in all_files:
if fnmatch.fnmatch(f, params.name_pattern):
basename = os.path.basename(f)
dfile = params.loc_output + os.path.sep + 'd_' + basename
ktable = params.loc_output + os.path.sep + 'k_' + basename
if os.path.exists(dfile) and os.path.exists(ktable):
nfiles += 1
filenames.append(f)
# Load the kernel tables
# Load the difference images into a data cube
print len(filenames), 'files found'
dates = np.zeros(nfiles)
seeing = np.zeros(nfiles)
roundness = np.zeros(nfiles)
bgnd = np.zeros(nfiles)
signal = np.zeros(nfiles)
norm_std = np.zeros(nfiles, dtype=np.float64)
diff_std = np.zeros(nfiles, dtype=np.float64)
n_kernel = np.zeros(nfiles, dtype=np.int32)
n_coeffs = np.zeros(nfiles, dtype=np.int32)
kindex_x = np.arange(0, dtype=np.int32)
kindex_y = np.arange(0, dtype=np.int32)
kindex_ext = np.arange(0, dtype=np.int32)
coeffs = np.arange(0, dtype=np.float64)
filenames.sort()
if not converge:
locate_iterations = 1
threshold = -10
for iteration in range(np.max([1, locate_iterations])):
ix0 = np.int32(x0 + 0.5)
iy0 = np.int32(y0 + 0.5)
x_patch = x0 - ix0 + patch_half_width
y_patch = y0 - iy0 + patch_half_width
patch_size = 2 * patch_half_width + 1
patch_slice = (ix0 - patch_half_width, ix0 + patch_half_width + 1,
iy0 - patch_half_width, iy0 + patch_half_width + 1)
d_image_stack = np.zeros((nfiles, patch_size, patch_size),
dtype=np.float64)
inv_var_image_stack = np.zeros((nfiles, patch_size, patch_size),
dtype=np.float64)
for i, f in enumerate(filenames):
basename = os.path.basename(f)
ktable = params.loc_output + os.path.sep + 'k_' + basename
kernelIndex, extendedBasis, c, params = IO.read_kernel_table(
ktable, params)
coeffs = np.hstack((coeffs, c))
kindex_x = np.hstack((kindex_x, kernelIndex[:, 0].T))
kindex_y = np.hstack((kindex_y, kernelIndex[:, 1].T))
kindex_ext = np.hstack((kindex_ext, extendedBasis))
n_kernel[i] = kernelIndex.shape[0]
n_coeffs[i] = c.shape[0]
dates[i] = IO.get_date(params.loc_data + os.path.sep + basename,
key=params.datekey) - 2450000
seeing[i], roundness[i], bgnd[i], signal[i] = IM.compute_fwhm(
f, params, width=20, image_name=True)
dfile = params.loc_output + os.path.sep + 'd_' + basename
nfile = params.loc_output + os.path.sep + 'n_' + basename
zfile = params.loc_output + os.path.sep + 'z_' + basename
diff, _ = IO.read_fits_file(dfile)
mask, _ = IO.read_fits_file(zfile)
diff_sc = IM.undo_photometric_scale(diff, c, params.pdeg)
diff_sc *= mask
d_image_stack[i, :, :] = diff_sc[patch_slice[2]:patch_slice[3],
patch_slice[0]:patch_slice[1]]
norm, _ = IO.read_fits_file(nfile, slice=patch_slice)
inv_var_image_stack[i, :, :] = (norm / d_image_stack[i, :, :])**2
diff_std[i] = np.std(diff)
d_image_stack[i, :, :] -= np.median(d_image_stack[i, :, :])
if save_stamps:
save_mosaic(
d_image_stack, nfiles, patch_size,
params.loc_output + os.path.sep + stamp_prefix + '.fits',
diff_std, threshold)
print 'kappa-clipping'
qd1 = np.arange(len(filenames))
#qd = np.where(diff_std[qd1]<10)
#qd1 = qd1[qd]
for iter in range(10):
qd = np.where(diff_std[qd1] < np.mean(diff_std[qd1]) +
(4.0 - 1.5 * (iter / 9.0)) * np.std(diff_std[qd1]))
qd1 = qd1[qd]
print iter, np.mean(diff_std[qd1]), np.std(diff_std[qd1]), np.mean(
diff_std[qd1]) + (4.0 - 3 *
(iter / 9.0)) * np.std(diff_std[qd1])
print 'mean(diff) :', np.mean(diff_std[qd1])
print 'std(diff) :', np.std(diff_std[qd1])
print '1-sig threshold:', np.mean(
diff_std[qd1]) + 1 * np.std(diff_std[qd1])
print '2-sig threshold:', np.mean(
diff_std[qd1]) + 2 * np.std(diff_std[qd1])
print '3-sig threshold:', np.mean(
diff_std[qd1]) + 3 * np.std(diff_std[qd1])
print '1-sig diff reject:', np.where(
diff_std > np.mean(diff_std[qd1]) + 1 * np.std(diff_std[qd1]))
print '2-sig diff reject:', np.where(
diff_std > np.mean(diff_std[qd1]) + 2 * np.std(diff_std[qd1]))
print '3-sig diff reject:', np.where(
diff_std > np.mean(diff_std[qd1]) + 3 * np.std(diff_std[qd1]))
threshold = np.mean(
diff_std[qd1]) + q_sigma_threshold * np.std(diff_std[qd1])
threshold2 = np.mean(diff_std[qd1]) + 2 * np.std(diff_std[qd1])
threshold3 = np.mean(diff_std[qd1]) + 3 * np.std(diff_std[qd1])
if locate_date_range is not None:
diff_std_copy = diff_std.copy()
diff_std = diff_std * 0.0 + 2 * threshold
pp = np.where((dates > locate_date_range[0])
& (dates < locate_date_range[1]))[0]
if pp.any():
diff_std[pp] = diff_std_copy[pp]
else:
print 'Error: No images found in date range', locate_date_range
print 'Reverting to all dates.'
diff_std = diff_std_copy
dsum = np.zeros((patch_size, patch_size), dtype=np.float64)
for i in range(nfiles):
if diff_std[i] < threshold3:
dsum += d_image_stack[i, :, :]
IO.write_image(
dsum, params.loc_output + os.path.sep + 'dsum%d.fits' % iteration)
dr = patch_half_width - int(locate_half_width)
dsum[:dr, :] = 0.0
dsum[-dr:, :] = 0.0
dsum[:, :dr] = 0.0
dsum[:, -dr:] = 0.0
ind_dsum_max = np.unravel_index(dsum.argmax(), dsum.shape)
print 'Iteration', iteration, ': dsum maximum located at ', ind_dsum_max
if locate and converge:
y0 += ind_dsum_max[0] - patch_half_width
x0 += ind_dsum_max[1] - patch_half_width
# Read the PSF
psf_image = params.loc_output + os.path.sep + 'psf.fits'
psf, psf_hdr = fits.getdata(psf_image, 0, header='true')
psf_height = psf_hdr['PSFHEIGH']
psf_sigma_x = psf_hdr['PAR1'] * 0.8493218
psf_sigma_y = psf_hdr['PAR2'] * 0.8493218
psf_x = psf_hdr['PSFX']
psf_y = psf_hdr['PSFY']
psf_size = psf.shape[1]
psf_fit_rad = params.psf_fit_radius
psf_parameters = np.array([
psf_size, psf_height, psf_sigma_x, psf_sigma_y, psf_x, psf_y,
psf_fit_rad, params.gain
]).astype(np.float64)
if params.psf_profile_type == 'gaussian':
psf_sigma_x = psf_hdr['PAR1'] * 0.8493218
psf_sigma_y = psf_hdr['PAR2'] * 0.8493218
psf_parameters = np.array([
psf_size, psf_height, psf_sigma_x, psf_sigma_y, psf_x, psf_y,
psf_fit_rad, params.gain
]).astype(np.float64)
profile_type = 0
elif params.psf_profile_type == 'moffat25':
print 'params.psf_profile_type moffat25 not working yet. Exiting.'
sys.exit(0)
psf_sigma_x = psf_hdr['PAR1']
psf_sigma_y = psf_hdr['PAR2']
psf_sigma_xy = psf_hdr['PAR3']
psf_parameters = np.array([
psf_size, psf_height, psf_sigma_x, psf_sigma_y, psf_x, psf_y,
psf_fit_rad, params.gain, psf_sigma_xy
]).astype(np.float64)
profile_type = 1
else:
print 'params.psf_profile_type undefined'
sys.exit(0)
psf_0 = psf.astype(np.float64).copy()
psf_xd = psf.astype(np.float64).copy() * 0.0
psf_yd = psf.astype(np.float64).copy() * 0.0
flux = np.zeros(nfiles, dtype=np.float64)
dflux = np.zeros(nfiles, dtype=np.float64)
x0_arr = np.atleast_1d(np.array([x0], dtype=np.float64))
y0_arr = np.atleast_1d(np.array([y0], dtype=np.float64))
cu_photom_converge(
profile_type, patch_half_width, params.pdeg, params.sdeg, nfiles,
n_kernel, kindex_x, kindex_y, kindex_ext, n_coeffs,
coeffs.astype(np.float64), psf_parameters, psf_0, psf_xd, psf_yd,
np.float64(d_image_stack.ravel()), inv_var_image_stack, diff_std,
np.float64(threshold), x0_arr, y0_arr, x_patch, y_patch, diff.shape[1],
diff.shape[0], 16, 16, flux, dflux, np.float64(params.gain),
np.int32(converge), np.float64(2.5))
if save_stamps:
save_mosaic(
d_image_stack, nfiles, patch_size,
params.loc_output + os.path.sep + 'p' + stamp_prefix + '.fits',
diff_std, threshold)
if locate_date_range is not None:
diff_std = diff_std_copy
return dates, seeing, roundness, bgnd, signal, flux, dflux, diff_std / threshold, x0_arr[
0], y0_arr[0]
def photom_variable_star(x0,y0,params,patch_half_width=15,converge=True,save_stamps=False,stamp_prefix='mosaic',locate=True,locate_iterations=2, locate_half_width=14,q_sigma_threshold=1.0,locate_date_range=None): from astropy.io import fits from scipy.ndimage.filters import median_filter outer_radius = 15 inner_radius = 12 diameter = 2*outer_radius + 1 x = np.arange(diameter)-outer_radius xx,yy = np.meshgrid(x,x) filter_kernel = np.zeros((diameter,diameter)) filter_kernel[xx**2+yy**2<=outer_radius**2] = 1 filter_kernel[xx**2+yy**2<=inner_radius**2] = 0 def save_mosaic(stack,nfiles,patch_size,name,diff_std,threshold): stamps_per_row = int(np.sqrt(nfiles)) nrows = (nfiles-1)/stamps_per_row+1; mx = stamps_per_row*(patch_size+1)+1 my = nrows*(patch_size+1)+1 mosaic = np.ones((my,mx))*1000.0 for i in range(nfiles): mosaic[(i/stamps_per_row)*(patch_size+1)+1:(i/stamps_per_row+1)*(patch_size+1), \ (i%stamps_per_row)*(patch_size+1)+1:(i%stamps_per_row+1)*(patch_size+1)] \ = stack[i,:,:] if diff_std[i] > threshold: mosaic[(i/stamps_per_row)*(patch_size+1)+1:(i/stamps_per_row+1)*(patch_size+1), \ (i%stamps_per_row)*(patch_size+1)+1] = -1000.0 mosaic[(i/stamps_per_row)*(patch_size+1)+1:(i/stamps_per_row+1)*(patch_size+1), \ (i%stamps_per_row+1)*(patch_size+1)-1] = -1000.0 mosaic[(i/stamps_per_row)*(patch_size+1)+1, \ (i%stamps_per_row)*(patch_size+1)+1:(i%stamps_per_row+1)*(patch_size+1)] = -1000.0 mosaic[(i/stamps_per_row+1)*(patch_size+1)-1, \ (i%stamps_per_row)*(patch_size+1)+1:(i%stamps_per_row+1)*(patch_size+1)] = -1000.0 IO.write_image(mosaic,name) # Obtain a list of files all_files = os.listdir(params.loc_data) all_files.sort() filenames = [] nfiles = 0 print 'Searching in', params.loc_output, 'for', params.name_pattern for f in all_files: if fnmatch.fnmatch(f,params.name_pattern): basename = os.path.basename(f) dfile = params.loc_output+os.path.sep+'d_'+basename ktable = params.loc_output+os.path.sep+'k_'+basename if os.path.exists(dfile) and os.path.exists(ktable): nfiles += 1 filenames.append(f) # Load the kernel tables # Load the difference images into a data cube print len(filenames), 'files found' dates = np.zeros(nfiles) seeing = np.zeros(nfiles) roundness = np.zeros(nfiles) bgnd = np.zeros(nfiles) signal = np.zeros(nfiles) norm_std = np.zeros(nfiles,dtype=np.float64) diff_std = np.zeros(nfiles,dtype=np.float64) n_kernel = np.zeros(nfiles,dtype=np.int32) n_coeffs = np.zeros(nfiles,dtype=np.int32) kindex_x = np.arange(0,dtype=np.int32) kindex_y = np.arange(0,dtype=np.int32) kindex_ext = np.arange(0,dtype=np.int32) coeffs = np.arange(0,dtype=np.float64) filenames.sort() if not converge: locate_iterations = 1 threshold = -10 for iteration in range(np.max([1,locate_iterations])): #ix0 = np.int32(x0+0.5) #iy0 = np.int32(y0+0.5) ix0 = np.int32(x0) iy0 = np.int32(y0) x_patch = x0 - ix0 + patch_half_width y_patch = y0 - iy0 + patch_half_width patch_size = 2*patch_half_width+1 patch_slice = np.array([ix0-patch_half_width, ix0+patch_half_width+1, iy0-patch_half_width, iy0+patch_half_width+1]) print 'patch_slice:', patch_slice # check that patch doesn't overlap the edge of the image f = filenames[0] diff, _ = IO.read_fits_file(params.loc_output+os.path.sep+'d_'+os.path.basename(f)) nx = diff.shape[1] ny = diff.shape[0] delta_patch_x = 0 delta_patch_y = 0 if patch_slice[0] < 0: delta_patch_x = -patch_slice[0] elif patch_slice[1] >= nx: delta_patch_x = nx - patch_slice[1] - 1 if patch_slice[2] < 0: delta_patch_y = -patch_slice[2] elif patch_slice[3] >= ny: delta_patch_y = ny - patch_slice[3] - 1 print 'delta_patch_x, delta_patch_y:', delta_patch_x, delta_patch_y patch_slice += np.array([delta_patch_x,delta_patch_x,delta_patch_y,delta_patch_y]) print 'patch_slice:', patch_slice x_patch -= delta_patch_x y_patch -= delta_patch_y d_image_stack = np.zeros((nfiles,patch_size,patch_size),dtype=np.float64) inv_var_image_stack = np.zeros((nfiles,patch_size,patch_size),dtype=np.float64) dmask = np.ones([patch_size,patch_size],dtype=np.bool) dmask_rad = 8.0 dmix = np.linspace(-patch_half_width,patch_half_width,patch_size) - delta_patch_x dmiy = np.linspace(-patch_half_width,patch_half_width,patch_size) - delta_patch_y dmx, dmy = np.meshgrid(dmix,dmiy,indexing='ij') dmask[dmx**2 + dmy**2 < dmask_rad**2] = False for i, f in enumerate(filenames): basename = os.path.basename(f) ktable = params.loc_output+os.path.sep+'k_'+basename kernelIndex, extendedBasis, c, params = IO.read_kernel_table(ktable,params) coeffs = np.hstack((coeffs,c)) kindex_x = np.hstack((kindex_x,kernelIndex[:,0].T)) kindex_y = np.hstack((kindex_y,kernelIndex[:,1].T)) kindex_ext = np.hstack((kindex_ext,extendedBasis)) n_kernel[i] = kernelIndex.shape[0] n_coeffs[i] = c.shape[0] dates[i] = IO.get_date(params.loc_data+os.path.sep+basename,key=params.datekey) if dates[i] > 2450000: dates[i] -= 2450000 seeing[i], roundness[i], bgnd[i], signal[i] = IM.compute_fwhm(f,params,width=20,image_name=True) dfile = params.loc_output+os.path.sep+'d_'+basename nfile = params.loc_output+os.path.sep+'n_'+basename zfile = params.loc_output+os.path.sep+'sm_'+basename ivfile = params.loc_output+os.path.sep+'iv_'+basename diff, _ = IO.read_fits_file(dfile) mask, _ = IO.read_fits_file(zfile) iv, _ = IO.read_fits_file(ivfile) diff_sc = IM.undo_photometric_scale(diff,c,params.pdeg) #diff_sc = diff #diff_sc -= median_filter(diff_sc,footprint=filter_kernel) diff_sc *= mask d_image_stack[i,:,:] = diff_sc[patch_slice[2]:patch_slice[3],patch_slice[0]:patch_slice[1]] inv_var_image_stack[i,:,:], _ = IO.read_fits_file(ivfile,slice=patch_slice) #inv_var_image_stack[i,:,:] = (norm / d_image_stack[i,:,:])**2 #diff_std[i] = np.std(diff) diff_std[i] = np.std(d_image_stack[i,:,:][dmask]) d_image_stack[i,:,:] -= np.median(d_image_stack[i,:,:]) print 'kappa-clipping' qd = np.arange(len(filenames)) qd1 = np.where(np.isfinite(diff_std))[0] for iter in range(10): qd = np.where(diff_std[qd1]<np.mean(diff_std[qd1])+(4.0-1.5*(iter/9.0))*np.std(diff_std[qd1]))[0] qd1 = qd1[qd] print iter, np.mean(diff_std[qd1]), np.std(diff_std[qd1]), np.mean(diff_std[qd1])+(4.0-3*(iter/9.0))*np.std(diff_std[qd1]) print 'mean(diff) :',np.mean(diff_std[qd1]) print 'std(diff) :',np.std(diff_std[qd1]) print '1-sig threshold:', np.mean(diff_std[qd1])+1*np.std(diff_std[qd1]) print '2-sig threshold:', np.mean(diff_std[qd1])+2*np.std(diff_std[qd1]) print '3-sig threshold:', np.mean(diff_std[qd1])+3*np.std(diff_std[qd1]) print '1-sig diff reject:',np.where(diff_std>np.mean(diff_std[qd1])+1*np.std(diff_std[qd1])) print '2-sig diff reject:',np.where(diff_std>np.mean(diff_std[qd1])+2*np.std(diff_std[qd1])) print '3-sig diff reject:',np.where(diff_std>np.mean(diff_std[qd1])+3*np.std(diff_std[qd1])) threshold = np.mean(diff_std[qd1])+q_sigma_threshold*np.std(diff_std[qd1]) threshold2 = np.mean(diff_std[qd1])+2*np.std(diff_std[qd1]) threshold3 = np.mean(diff_std[qd1])+3*np.std(diff_std[qd1]) if locate_date_range is not None: diff_std_copy = diff_std.copy() diff_std = diff_std*0.0 + 100.0*threshold pp = np.where((dates>locate_date_range[0]) & (dates<locate_date_range[1]))[0] if pp.any(): print 'Using images ',pp diff_std[pp] = diff_std_copy[pp] else: print 'Error: No images found in date range',locate_date_range print 'Reverting to all dates.' diff_std = diff_std_copy print 'zeros:' for i in range(nfiles): print i, np.sum(np.abs(d_image_stack[i,:,:]) < 1.e-6) if np.isnan(inv_var_image_stack[i,:,:]).any() or np.sum(np.abs(d_image_stack[i,:,:]) < 1.e-6) > 5: diff_std[i] = 100.0*threshold inv_var_image_stack[i,:,:] = inv_var_image_stack[i,:,:]*0.0 if save_stamps: save_mosaic(d_image_stack,nfiles,patch_size,params.loc_output+os.path.sep+stamp_prefix+'.fits',diff_std,threshold) dsum = np.zeros((patch_size,patch_size),dtype=np.float64) for i in range(nfiles): if diff_std[i] < threshold3: dsum += d_image_stack[i,:,:] IO.write_image(dsum,params.loc_output+os.path.sep+'dsum%d.fits'%iteration) dr = patch_half_width-int(locate_half_width) print 'dr:', dr dsum[:dr-delta_patch_y,:] = 0.0 dsum[-dr-delta_patch_y:,:] = 0.0 dsum[:,:dr-delta_patch_x] = 0.0 dsum[:,-dr-delta_patch_x:] = 0.0 IO.write_image(dsum,params.loc_output+os.path.sep+'dsum_m%d.fits'%iteration) ind_dsum_max = np.unravel_index(dsum.argmax(),dsum.shape) print 'Iteration',iteration,': dsum maximum located at ',ind_dsum_max if locate and converge: y0 += ind_dsum_max[0] - patch_half_width + delta_patch_y x0 += ind_dsum_max[1] - patch_half_width + delta_patch_x # Read the PSF psf_image = params.loc_output+os.path.sep+'psf.fits' psf,psf_hdr = fits.getdata(psf_image,0,header='true') psf_height = psf_hdr['PSFHEIGH'] psf_sigma_x = psf_hdr['PAR1']*0.8493218 psf_sigma_y = psf_hdr['PAR2']*0.8493218 psf_x = psf_hdr['PSFX'] psf_y = psf_hdr['PSFY'] psf_size = psf.shape[1] psf_fit_rad = params.psf_fit_radius psf_parameters = np.array([psf_size,psf_height,psf_sigma_x,psf_sigma_y,psf_x, psf_y,psf_fit_rad,params.gain]).astype(np.float64) if params.psf_profile_type == 'gaussian': psf_sigma_x = psf_hdr['PAR1']*0.8493218 psf_sigma_y = psf_hdr['PAR2']*0.8493218 psf_parameters = np.array([psf_size,psf_height,psf_sigma_x,psf_sigma_y,psf_x, psf_y,psf_fit_rad,params.gain]).astype(np.float64) profile_type = 0 elif params.psf_profile_type == 'moffat25': print 'params.psf_profile_type moffat25 not working yet. Exiting.' sys.exit(0) psf_sigma_x = psf_hdr['PAR1'] psf_sigma_y = psf_hdr['PAR2'] psf_sigma_xy = psf_hdr['PAR3'] psf_parameters = np.array([psf_size,psf_height,psf_sigma_x,psf_sigma_y,psf_x, psf_y, psf_fit_rad,params.gain,psf_sigma_xy]).astype(np.float64) profile_type = 1 else: print 'params.psf_profile_type undefined' sys.exit(0) psf_0 = psf.astype(np.float64).copy() psf_xd = psf.astype(np.float64).copy()*0.0 psf_yd = psf.astype(np.float64).copy()*0.0 flux = np.zeros(nfiles,dtype=np.float64) dflux = np.zeros(nfiles,dtype=np.float64) x0_arr = np.atleast_1d(np.array([x0],dtype=np.float64)) y0_arr = np.atleast_1d(np.array([y0],dtype=np.float64)) print 'Converging photometry' print 'x0, y0:', x0, y0 print 'x_patch, y_patch:', x_patch, y_patch good_images = np.where(diff_std < threshold)[0] print 'using images', good_images print 'threshold', threshold for i, f in enumerate(filenames): if i in good_images: print i, 'd_'+f, diff_std[i] cu_photom_converge(profile_type, patch_half_width, params.pdeg, params.sdeg, nfiles, n_kernel, kindex_x, kindex_y, kindex_ext, n_coeffs, coeffs.astype(np.float64), psf_parameters, psf_0, psf_xd, psf_yd, np.float64(d_image_stack.ravel()), np.float64(inv_var_image_stack.ravel()), diff_std, np.float64(threshold), x0_arr, y0_arr, x_patch, y_patch, diff.shape[1], diff.shape[0], 16, 16, flux, dflux, np.float64(params.gain),np.int32(converge),np.float64(2.5)) if save_stamps: save_mosaic(d_image_stack,nfiles,patch_size,params.loc_output+os.path.sep+'p'+stamp_prefix+'.fits',diff_std,threshold) if locate_date_range is not None: diff_std = diff_std_copy return dates, seeing, roundness, bgnd, signal, flux, dflux, diff_std/threshold, x0_arr[0], y0_arr[0]