def makemap(band,skypos,trange,skyrange,response=False,verbose=0,detsize=1.1):
imsz = gxt.deg2pix(skypos,skyrange)
photons = np.array(gQuery.getArray(gQuery.skyrect(band,
skypos[0],skypos[1],trange[0],trange[1],skyrange[0],skyrange[1]),
verbose=verbose),dtype='float64')
try:
events = {'t':photons[:,0 ]/tscale,'ra':photons[:,1],'dec':photons[:,2],
'xi':photons[:,3],'eta':photons[:,4],
'x':photons[:,5], 'y':photons[:,6]}
except IndexError:
if verbose>2:
print 'No events found at {s} +/- {r} in {t}.'.format(
s=skypos,r=skyrange,t=trange)
return np.zeros(imsz)
# Trim the data on detsize
col, row = ct.xieta2colrow(events['xi'],events['eta'],band)
ix = np.where((1.25/800.)*mc.distance(col,row,400,400)<=detsize)
n = len(ix[0])
m = len(col)
#print 'With detsize {d} using {n} of {m} data.'.format(d=detsize,n=n,m=m)
if n == 0:
return np.zeros(imsz)
for k in events.keys():
events[k] = events[k][ix]
events = ct.hashresponse(band,events)
wcs = define_wcs(skypos,skyrange,width=False,height=False)
coo = zip(events['ra'],events['dec'])
foc = wcs.sip_pix2foc(wcs.wcs_world2pix(coo,1),1)
weights = 1./events['response'] if response else None
H,xedges,yedges=np.histogram2d(foc[:,1]-0.5,foc[:,0]-0.5,bins=imsz,
range=([ [0,imsz[0]],[0,imsz[1]] ]),weights=weights)
return H
def integrate_map(band,skypos,tranges,skyrange,width=False,height=False,
verbose=0,memlight=False,hdu=False,retries=20,response=False,
detsize=1.1):
""" Integrate an image over some number of time ranges. Use a reduced
memory optimization (at the expense of more web queries) if requested.
"""
imsz = gxt.deg2pix(skypos,skyrange)
img = np.zeros(imsz)
for trange in tranges:
# If memlight is requested, break the integration into
# smaller chunks.
# FIXME: memlight gives slightly wrong answers right now
# This is probably due to a quirk of SQL, per issue #140.
# Deprecating memlight until this can be resolved.
step = memlight if memlight else trange[1]-trange[0]
for i in np.arange(trange[0],trange[1],step):
t0,t1=i,i+step
if verbose:
mc.print_inline('Coadding '+str(t0)+' to '+str(t1))
img += makemap(band,skypos,[t0,t1],skyrange,response=response,
verbose=verbose,detsize=detsize)
if response: # This is an intensity map.
img /= dbt.compute_exptime(band,trange,skypos=skypos,
verbose=verbose)
return img
def sigmaclip_bg(data,radius,annulus,skypos,maxiter=10,sigmaclip=3.,
gausslim=50.,verbose=0,pixsz=0.000416666666666667):
"""Produce an estimate of background counts within an aperture (radius)
using a sigma clipping method for extracting the background from an
annulus.
This attempts to reproduce the calcuations of the backcalc() function in
mapaps/poissonbg.c of the mission pipeline. (Probably written by Ted Wyder.)
"""
# FIXME: Does not apply response!
# This cut is now handled by the ugly loop below, which barely dodges a
# conceptula issue about fractional pixels...
#ix = np.where((d>annulus[0]) & (d<annulus[1]))
imsz=gxt.deg2pix(skypos,[annulus[1]*2,annulus[1]*2])
wcs=define_wcs(skypos,[annulus[1]*2,annulus[1]*2])
foc_ra,foc_dec=wcs.sip_pix2foc(wcs.wcs_world2pix(data['ra'],data['dec'],1),1)
H,xedges,yedges=np.histogram2d(foc_ra-0.5,foc_dec-0.5,bins=imsz,
range=([ [0,imsz[0]],[0,imsz[1]]]))
# Convert Gaussian sigma to a probability
problim = 0.5*scipy.special.erfc(sigmaclip/np.sqrt(2.0))
# Mask out non-annulus regions... there's probalby a more pythonic way
bgimg=np.copy(H)
for i in range(H.shape[0]):
for j in range(H.shape[1]):
# Add a little buffer to account for pixel widths?
# FIXME? including everything within the annulus...
# if (mc.distance(H.shape[0]/2.,H.shape[1]/2.,i,j)<annulus[0]/pixsz or
if mc.distance(H.shape[0]/2.,H.shape[1]/2.,i,j)>annulus[1]/pixsz:#):
bgimg[i,j]=-1
ix=np.where(bgimg>=0)
m,s=bgimg[ix].mean(),bgimg[ix].std()
d = 1.
for i in range(maxiter):
if d<=10e-5 or m<2:
continue
if m>=gausslim:
# Mask anything outside of 3 sigma from the mean (of unmasked data)
klim=m+sigmaclip*np.sqrt(m)#s
klo=m-sigmaclip*np.sqrt(m)#s
if verbose:
print 'Gaussian cut: {klo} to {klim}'.format(klo=klo,klim=klim)
else:
klim = scipy.special.gammainccinv(m,problim)
klo = -1 # None
if verbose:
print 'Poisson cut: {klo} to {klim}'.format(klo=klo,klim=klim)
ix = np.where((bgimg>=klim) | (bgimg<=klo))
bgimg[ix]=-1
ix=np.where(bgimg>=0)
d = np.abs((bgimg[ix].mean()-m)/m)# - 1)
m,s=bgimg[ix].mean(),bgimg[ix].std()
ix = np.where(bgimg>=0)
return mc.area(radius)*bgimg[ix].mean()/mc.area(pixsz)
def rrhr(band,skypos,tranges,skyrange,width=False,height=False,stepsz=1.,
verbose=0,calpath='../cal/',tscale=1000.,response=True,hdu=False,
retries=20):
"""Generate a high resolution relative response (rrhr) map."""
imsz = gxt.deg2pix(skypos,skyrange)
# TODO the if width / height
flat = get_fits_data(flat_filename(band,calpath),verbose=verbose)
flatinfo = get_fits_header(flat_filename(band,calpath))
npixx,npixy = flat.shape
fltsz = flat.shape
pixsz = flatinfo['CDELT2']
detsize = 1.25
# Rotate the flat into the correct orientation to start.
flat = np.flipud(np.rot90(flat))
# NOTE: This upsample interpolation is done _last_ in the canonical
# pipeline as part of the poissonbg.c routine.
# The interpolation function is "congrid" in the same file.
# TODO: Should this be first order interpolation? (i.e. bilinear)
hrflat = scipy.ndimage.interpolation.zoom(flat,4.,order=0,prefilter=False)
img = np.zeros(hrflat.shape)[
hrflat.shape[0]/2.-imsz[0]/2.:hrflat.shape[0]/2.+imsz[0]/2.,
hrflat.shape[1]/2.-imsz[1]/2.:hrflat.shape[1]/2+imsz[1]/2.]
for trange in tranges:
t0,t1=trange
entries = gQuery.getArray(gQuery.aspect(t0,t1),retries=retries)
n = len(entries)
asptime = np.float64(np.array(entries)[:,2])/tscale
aspra = np.float32(np.array(entries)[:,3])
aspdec = np.float32(np.array(entries)[:,4])
asptwist= np.float32(np.array(entries)[:,5])
aspflags= np.float32(np.array(entries)[:,6])
asptwist= np.float32(np.array(entries)[:,9])
aspra0 = np.zeros(n)+skypos[0]
aspdec0 = np.zeros(n)+skypos[1]
xi_vec, eta_vec = gnomonic.gnomfwd_simple(
aspra,aspdec,aspra0,aspdec0,-asptwist,1.0/36000.,0.)
col = 4.*( ((( xi_vec/36000.)/(detsize/2.)*(detsize/(fltsz[0]*pixsz)) + 1.)/2. * fltsz[0]) - (fltsz[0]/2.) )
row = 4.*( (((eta_vec/36000.)/(detsize/2.)*(detsize/(fltsz[1]*pixsz)) + 1.)/2. * fltsz[1]) - (fltsz[1]/2.) )
vectors = rotvec(np.array([col,row]),-asptwist)
for i in range(n):
if verbose>1:
print_inline('Stamping '+str(asptime[i]))
# FIXME: Clean this mess up a little just for clarity.
img += scipy.ndimage.interpolation.shift(scipy.ndimage.interpolation.rotate(hrflat,-asptwist[i],reshape=False,order=0,prefilter=False),[vectors[1,i],vectors[0,i]],order=0,prefilter=False)[hrflat.shape[0]/2.-imsz[0]/2.:hrflat.shape[0]/2.+imsz[0]/2.,hrflat.shape[1]/2.-imsz[1]/2.:hrflat.shape[1]/2+imsz[1]/2.]*dbt.compute_exptime(band,[asptime[i],asptime[i]+1],verbose=verbose,retries=retries)*gxt.compute_flat_scale(asptime[i]+0.5,band,verbose=0)
return img
def define_wcs(skypos,skyrange,width=False,height=False,verbose=0,
pixsz=0.000416666666666667):
"""Define the world coordinate system (WCS)."""
if verbose:
print_inline('Defining World Coordinate System (WCS).')
wcs = pywcs.WCS(naxis=2) # NAXIS = 2
imsz = gxt.deg2pix(skypos,skyrange)
wcs.wcs.cdelt = np.array([-pixsz,pixsz])
wcs.wcs.ctype = ['RA---TAN','DEC--TAN']
wcs.wcs.crpix = [(imsz[1]/2.)+0.5,(imsz[0]/2.)+0.5]
wcs.wcs.crval = skypos
return wcs
def makemap(band,skypos,trange,skyrange,response=False,verbose=0):
imsz = gxt.deg2pix(skypos,skyrange)
photons = np.array(gQuery.getArray(gQuery.skyrect(band,
skypos[0],skypos[1],trange[0],trange[1],skyrange[0],skyrange[1]),
verbose=verbose),dtype='float64')
events = {'t':photons[:,0 ]/tscale, 'ra':photons[:,1], 'dec':photons[:,2],
'xi':photons[:,3],'eta':photons[:,4],
'x':photons[:,5], 'y':photons[:,6]}
if len(events['t'])==0:
return np.zeros(imsz)
events = ct.hashresponse(band,events)
wcs = define_wcs(skypos,skyrange,width=False,height=False)
coo = zip(events['ra'],events['dec'])
foc = wcs.sip_pix2foc(wcs.wcs_world2pix(coo,1),1)
weights = 1./events['response'] if response else None
H,xedges,yedges=np.histogram2d(foc[:,1]-0.5,foc[:,0]-0.5,bins=imsz,
range=([ [0,imsz[0]],[0,imsz[1]] ]),weights=weights)
return H
def countmap(band,skypos,tranges,skyrange,width=False,height=False,
verbose=0,tscale=1000.,memlight=False,hdu=False,retries=20):
"""Create a count (cnt) map."""
imsz = gxt.deg2pix(skypos,skyrange)
count = np.zeros(imsz)
for trange in tranges:
# If memlight is requested, break the integration into
# smaller chunks.
step = memlight if memlight else trange[1]-trange[0]
for i in np.arange(trange[0],trange[1],step):
t0,t1=i,i+step
if verbose:
print_inline('Coadding '+str(t0)+' to '+str(t1))
events = gQuery.getArray(gQuery.rect(band,skypos[0],skypos[1],t0,t1,
skyrange[0],skyrange[1]),
verbose=verbose,retries=retries)
# Check that there is actually data here.
if not events:
if verbose>1:
print "No data in "+str([t0,t1])
continue
times = np.array(events,dtype='float64')[:,0 ]/tscale
coo = np.array(events,dtype='float64')[:,1:]
# If there's no data, return a blank image.
if len(coo)==0:
if verbose:
print 'No data in this frame: '+str([t0,t1])
continue
# Define World Coordinate System (WCS)
wcs = define_wcs(skypos,skyrange,width=False,height=False)
# Map the sky coordinates onto the focal plane
foc = wcs.sip_pix2foc(wcs.wcs_world2pix(coo,1),1)
# Bin the events into actual image pixels
H,xedges,yedges=np.histogram2d(foc[:,1]-0.5,foc[:,0]-0.5,
bins=imsz,range=([ [0,imsz[0]],[0,imsz[1]] ]))
count += H
return count
def test_deg2pix_pole(self):
skypos,skyrange=[90.,90.],[1.,1.]
pix = gt.deg2pix(skypos,skyrange)
self.assertAlmostEqual(pix[0],2400.)
self.assertAlmostEqual(pix[1],21.)
def test_deg2pix_equator(self):
skypos,skyrange=[0.,0.],[1.,1.]
pix = gt.deg2pix(skypos,skyrange)
self.assertAlmostEqual(pix[0],2400.)
self.assertAlmostEqual(pix[1],2400.)
