def set_imagepair(self, val):
if not self.theapp.current_target:
popup = WarningDialog(text='You need to select a target (on the Observing Screen) before proceeding!')
popup.open()
return
self.pair_index = self.pairstrings.index(val)
fitsfile = self.paths['out']+re.sub(' ','',re.sub('.fits','',val))+'.fits'
if not path.isfile(fitsfile):
popup = WarningDialog(text='You have to select an extraction'\
'region for this image pair \nbefore you can move on to this step.')
popup.open()
return
self.current_impair = FitsImage(fitsfile)
self.region = self.current_impair.get_header_keyword(*('EXREG' + x for x in ['X1','Y1','X2','Y2']))
if not any(self.region):
popup = WarningDialog(text='You have to select an extraction'\
'region for this image pair \nbefore you can move on to this step.')
popup.open()
return
self.current_impair.load()
idata = ''.join(map(chr,self.current_impair.scaled))
self.itexture.blit_buffer(idata, colorfmt='luminance', bufferfmt='ubyte', \
size = self.current_impair.dimensions)
self.trace_axis = 0 if get_tracedir(self.current_target.instrument_id) == 'vertical' else 1
#tmp = self.current_impair.data_array[self.region[1]:self.region[3]+1,self.region[0]:self.region[2]+1]
#self.extractregion = med_normal(tmp)
tmp = self.theapp.extract_pairs[self.pair_index]
if self.trace_axis:
#self.trace_axis = 1
reg = [self.region[x] for x in [1, 0, 3, 2]]
self.extractregion = make_region(tmp[0], tmp[1], reg, self.current_flats)#.transpose()
else:
self.extractregion = make_region(tmp[0], tmp[1], self.region, self.current_flats).transpose()
reg = self.region[:]
reg[2] = reg[2] - reg[0]
reg[3] = reg[3] - reg[1]
self.iregion = self.itexture.get_region(*reg)
dims = [[0,0],list(self.extractregion.shape)]
dims[0][self.trace_axis] = 0.4 * self.extractregion.shape[self.trace_axis]
dims[1][self.trace_axis] = 0.6 * self.extractregion.shape[self.trace_axis]
self.tracepoints = robm(self.extractregion[dims[0][0]:dims[1][0]+1,\
dims[0][1]:dims[1][1]+1], axis = self.trace_axis)
self.tplot.points = interp_nan(list(enumerate(self.tracepoints)))
pxs, self.tracepoints = zip(*self.tplot.points)
self.drange = minmax(self.tracepoints)
self.ids.the_graph.add_plot(self.tplot)
def find_peaks(idata, npeak = 1, tracedir = None, pn = 'pos'):
data = np.array(idata) #make sure we're dealing with an array
if len(data.shape) > 1: #check for 2D array
if traceder is None:
tracedir = 1 #assume that the second axis is the right one...
#if idata is 2D, compress along the trace using a robust mean
data = robm(data, axis = tracedir)
#since argrelextrema isn't working, just return argmax
if pn == 'pos':
return np.nanargmax(data)
else:
return np.nanargmin(data)
ps = zip(range(data.size), data)
junk, data = zip(*interp_nan(ps)) #remove nans via interpolation
data = medfilt(data, 5) #a little smoothness never hurt
#find /all/ rel extrema:
maxima = argrelextrema(data, posneg[pn])
max_val = data[maxima]
priority = np.argsort(-np.fabs(max_val))
return maxima[priority[:npeak]]
def draw_trace(idata, x_val, pfit, nfit, fixdistort = False, fitdegree = 2, ptype = 'Gaussian'):
'''move along the trace axis, fitting each position with a model of the PSF'''
#pdb.set_trace()
ns = idata.shape[1]
midpoint = ns/2
tc1, tc2 = midpoint, midpoint + 1
fitp = deconstruct_composite(pfit)
fitn = deconstruct_composite(nfit)
p_amp, p_mean, p_sig = get_individual_params(*fitp)
n_amp, n_mean, n_sig = get_individual_params(*fitn)
n_p = len(p_mean)
n_n = len(n_mean)
#back-convert the custom model into a composite model
#fitp = build_composite(pfit, ptype)
#fitn = build_composite(nfit, ptype)
#trace = {'pos':[np.zeros(idata.shape) for _ in fitp._transforms], \
# 'neg':[np.zeros(idata.shape) for _ in fitn._transforms]}
#apertures = {'pos':[range(ns) for _ in fitp._transforms], \
# 'neg':[range(ns) for _ in fitn._transforms]}
trace = {'pos':[np.zeros(idata.shape) for _ in p_mean], \
'neg':[np.zeros(idata.shape) for _ in n_mean]}
apertures = {'pos':[range(ns) for _ in p_mean], \
'neg':[range(ns) for _ in n_mean]}
pcur1, ncur1 = deepcopy(fitp), deepcopy(fitn)
pcur2, ncur2 = deepcopy(fitp), deepcopy(fitn)
pmodel, nmodel = multi_peak_model(ptype, n_p), multi_peak_model(ptype, n_n)
#pcur1, ncur1 = deepcopy(pfit), deepcopy(nfit)
#pcur2, ncur2 = deepcopy(pfit), deepcopy(nfit)
down, up = True, True
#set up initial data for use with cross-correlation
piece0 = robm(idata[:, (max(tc1 - 20, 0), min(tc1 + 20, ns - 1))], axis=1)
junk, piece0 = zip(*interp_nan(list(enumerate(piece0))))
med = np.median(piece0)
p0 = np.clip(piece0, a_min = med, a_max = np.nanmax(piece0))
n0 = np.clip(piece0, a_min = np.nanmin(piece0), a_max = med)
while down or up:
#work in both directions from the middle
if tc1 >= 0:
lb = max(tc1 - 20, 0)
ub = min(tc1 + 20, ns-1)
piece = robm(idata[:,(lb,ub)], axis=1)
junk, piece = zip(*interp_nan(list(enumerate(piece))))
med = np.median(piece)
pdata = np.clip(piece,a_min=med,a_max=np.nanmax(piece))
ndata = np.clip(piece,a_min=np.nanmin(piece),a_max=med)
if pcur1 is not None:
offset = offset1d(p0, pdata)
pcur1 = [np.array([interp1d(x_val, pdata, kind='cubic', bounds_error=False)(f[1] + \
offset), f[1] + offset, f[2]]) for f in fitp]
#pnew1 = fitmethod(pcur1, x_val, pdata)
pnew1, psig1 = curve_fit(pmodel, x_val, pdata, pcur1)
pnmodel = build_composite(pnew1, ptype)
#for i, transform in enumerate(pnew1._transforms):
for i, transform in enumerate(pnmodel._transforms):
trace['pos'][i][:,tc1] = transform(x_val)
apertures['pos'][i][tc1] = transform.mean
pcur1 = pnew1
#print tc1, pcur1
if ncur1 is not None:
offset = offset1d(n0, ndata)
ncur1 = [np.array([interp1d(x_val, ndata, kind='cubic', bounds_error=False)(f[1] + \
offset), f[1] + offset, f[2]]) for f in fitn]
#nnew1 = fitmethod(ncur1, x_val, ndata)
nnew1, nsig1 = curve_fit(nmodel, x_val, ndata, ncur1)
nnmodel = build_composite(nnew1, ptype)
#for i, transform in enumerate(nnew1._transforms):
for i, transform in enumerate(nnmodel._transforms):
trace['neg'][i][:,tc1] = transform(x_val)
apertures['neg'][i][tc1] = transform.mean
ncur1 = nnew1
#print tc1, ncur1
tc1 -= 1
else:
down = False
if tc2 < ns:
lb = max(tc2 - 20, 0)
ub = min(tc2 + 20, ns-1)
piece = robm(idata[:,(lb,ub)], axis=1)
junk, piece = zip(*interp_nan(list(enumerate(piece))))
med = np.median(piece)
pdata = np.clip(piece,a_min=med,a_max=np.nanmax(piece))
ndata = np.clip(piece,a_min=np.nanmin(piece),a_max=med)
if pcur2 is not None:
offset = offset1d(p0, pdata)
pcur2 = [np.array([interp1d(x_val, pdata, kind='cubic', bounds_error=False)(f[1] + \
offset), f[1] + offset, f[2]]) for f in fitp]
#pnew2 = fitmethod(pcur2, x_val, pdata)
pnew2, psig2 = curve_fit(pmodel, x_val, pdata, pcur2)
pnmodel = build_composite(pnew2, ptype)
#for i, transform in enumerate(pnew2._transforms):
for i, transform in enumerate(pnmodel._transforms):
trace['pos'][i][:,tc2] = transform(x_val)
apertures['pos'][i][tc2] = transform.mean
pcur2 = pnew2
#print tc2, pcur2
if ncur2 is not None:
offset = offset1d(n0, ndata)
ncur2 = [np.array([interp1d(x_val, ndata, kind='cubic', bounds_error=False)(f[1] + \
offset), f[1] + offset, f[2]]) for f in fitn]
#nnew2 = fitmethod(ncur2, x_val, ndata)
nnew2, nsig2 = curve_fit(nmodel, x_val, ndata, ncur2)
nnmodel = build_composite(nnew2, ptype)
#for i, transform in enumerate(nnew2._transforms):
for i, transform in enumerate(nnmodel._transforms):
trace['neg'][i][:,tc2] = transform(x_val)
apertures['neg'][i][tc2] = transform.mean
ncur2 = nnew2
#print tc2, ncur2
tc2 += 1
else:
up = False
#import shelve
#f = shelve.open('/Users/gray/Desktop/trace-shelve')
#f['pos'] = trace['pos']
#f['neg'] = trace['neg']
if not fixdistort:
return trace
#pdb.set_trace()
if pcur1 is not None:
#identify the various aperture traces, subtract off the
#median x-position of each one, determine the median offset
#across apertures, and fit with a polynomial
if len(apertures['pos']) > 1:
ap = np.array(zip(*apertures['pos'])).squeeze()
ns, nap = ap.shape
meds = np.median(ap, axis=1)
meds = np.repeat(meds.reshape(ns, 1), nap, axis=1)
else:
ap = np.array(apertures['pos'][0]).squeeze()
ns, nap = ap.size, 1
meds = np.median(ap)
ap -= meds
off_x = np.median(ap, axis=0) if nap > 1 else ap
pinit = poly.Polynomial1D(fitdegree)
x_trace = np.arange(ns)
posfit = fitmethod(pinit, x_trace, off_x)
posfit
else: posfit = None
if ncur1 is not None:
if len(apertures['neg']) > 1:
ap = np.array(zip(*apertures['neg'])).squeeze()
ns, nap = ap.shape
meds = np.median(ap, axis=1)
meds = np.repeat(meds.reshape(ns, 1), nap, axis=1)
else:
ap = np.array(apertures['neg'][0]).squeeze()
ns, nap = ap.size, 1
meds = np.median(ap)
ap -= meds
off_x = np.median(ap, axis=0) if nap > 1 else ap
pinit = poly.Polynomial1D(fitdegree)
x_trace = np.arange(ns)
negfit = fitmethod(pinit, x_trace, off_x)
else: negfit = None
return posfit, negfit
