def main(argv):
try:
File = argv[0]
atr = readfile.read_attribute(File)
epoch = argv[1]
except:
usage()
sys.exit(1)
try:
outFile = argv[2]
except:
outFile = None
# Calculate look angle
pbase = ut.perp_baseline_timeseries(atr, dimension=1)
if pbase.shape[1] == 1:
print pbase
return pbase
k = atr['FILE_TYPE']
width = int(atr['WIDTH'])
length = int(atr['FILE_LENGTH'])
h5 = h5py.File(File, 'r')
epochList = sorted(h5[k].keys())
epoch = ptime.yyyymmdd(epoch)
epoch_idx = epochList.index(epoch)
pbase_y = pbase[epoch_idx, :].reshape(length, 1)
pbase_xy = np.tile(pbase_y, (1, width))
if not outFile:
outFile = 'perpBaseline_' + epoch + '.h5'
print 'writing >>> ' + outFile
atr['FILE_TYPE'] = 'mask'
atr['UNIT'] = 'm'
writefile.write(pbase_xy, atr, outFile)
return outFile
def main(argv):
inps = cmdLineParse()
suffix = '_demErr'
if not inps.outfile:
inps.outfile = os.path.splitext(
inps.timeseries_file)[0] + suffix + os.path.splitext(
inps.timeseries_file)[1]
# 1. template_file
if inps.template_file:
print 'read option from template file: ' + inps.template_file
inps = read_template2inps(inps.template_file, inps)
# Read Time Series
print "loading time series: " + inps.timeseries_file
atr = readfile.read_attribute(inps.timeseries_file)
length = int(atr['FILE_LENGTH'])
width = int(atr['WIDTH'])
h5 = h5py.File(inps.timeseries_file)
date_list = sorted(h5['timeseries'].keys())
date_num = len(date_list)
print 'number of acquisitions: ' + str(date_num)
# Exclude date info
#inps.ex_date = ['20070115','20100310']
if inps.ex_date:
inps = get_exclude_date(inps, date_list)
if inps.ex_date:
inps.ex_flag = np.array([i not in inps.ex_date for i in date_list])
timeseries = np.zeros((len(date_list), length * width), np.float32)
prog_bar = ptime.progress_bar(maxValue=date_num, prefix='loading: ')
for i in range(date_num):
date = date_list[i]
d = h5['timeseries'].get(date)[:]
timeseries[i][:] = d.flatten('F')
prog_bar.update(i + 1, suffix=date)
del d
h5.close()
prog_bar.close()
# Perpendicular Baseline
print 'read perpendicular baseline'
try:
inps.pbase = ut.perp_baseline_timeseries(atr, dimension=0)
if inps.pbase.shape[1] > 1:
print '\tconsider P_BASELINE variation in azimuth direction'
else:
pbase = inps.pbase
except:
print '\tCannot find P_BASELINE_TIMESERIES from timeseries file.'
print '\tTrying to calculate it from interferograms file'
if inps.ifgram_file:
inps.pbase = np.array(
ut.perp_baseline_ifgram2timeseries(
inps.ifgram_file)[0]).reshape(date_num, 1)
else:
message = 'No interferogram file input!\n'+\
'Can not correct for DEM residula without perpendicular base info!'
raise Exception(message)
# Temporal Baseline
print 'read temporal baseline'
inps.tbase = np.array(ptime.date_list2tbase(date_list)[0]).reshape(
date_num, 1)
# Incidence angle (look angle in the paper)
if inps.incidence_angle:
if os.path.isfile(inps.incidence_angle):
print 'reading incidence angle from file: ' + inps.incidence_angle
inps.incidence_angle = readfile.read(inps.incidence_angle)[0]
else:
try:
inps.incidence_angle = np.array(float(inps.incidence_angle))
print 'use input incidence angle : ' + str(
inps.incidence_angle)
except:
raise ValueError('Can not read input incidence angle: ' +
str(inps.incidence_angle))
else:
print 'calculate incidence angle using attributes of time series file'
if inps.pbase.shape[1] > 1:
inps.incidence_angle = ut.incidence_angle(atr, dimension=2)
else:
inps.incidence_angle = ut.incidence_angle(atr, dimension=1)
inps.incidence_angle *= np.pi / 180.0
# Range distance
if inps.range_dis:
if os.path.isfile(inps.range_dis):
print 'reading range distance from file: ' + inps.range_dis
inps.range_dis = readfile.read(inps.range_dis)[0]
else:
try:
inps.range_dis = np.array(float(inps.range_dis))
print 'use input range distance : ' + str(inps.range_dis)
except:
raise ValueError('Can not read input incidence angle: ' +
str(inps.range_dis))
else:
print 'calculate range distance using attributes from time series file'
if inps.pbase.shape[1] > 1:
inps.range_dis = ut.range_distance(atr, dimension=2)
else:
inps.range_dis = ut.range_distance(atr, dimension=1)
# Design matrix - temporal deformation model using tbase
print '-------------------------------------------------'
if inps.phase_velocity:
print 'using phase velocity history'
A1 = np.ones((date_num - 1, 1))
A2 = (inps.tbase[1:date_num] + inps.tbase[0:date_num - 1]) / 2.0
A3 = (inps.tbase[1:date_num]**3 - inps.tbase[0:date_num - 1]**
3) / np.diff(inps.tbase, axis=0) / 6.0
#A3 = (inps.tbase[1:date_num]**2 + inps.tbase[1:date_num]*inps.tbase[0:date_num-1] +\
# inps.tbase[0:date_num-1]**2) / 6.0
else:
print 'using phase history'
A1 = np.hstack((np.ones((date_num, 1)), inps.tbase))
A2 = inps.tbase**2 / 2.0
A3 = inps.tbase**3 / 6.0
# Polynomial order of model
print "temporal deformation model's polynomial order = " + str(
inps.poly_order)
if inps.poly_order == 1: A_def = A1
elif inps.poly_order == 2: A_def = np.hstack((A1, A2))
elif inps.poly_order == 3: A_def = np.hstack((A1, A2, A3))
# step function
if inps.step_date:
print "temporal deformation model's step function step at " + inps.step_date
step_yy = ptime.yyyymmdd2years(inps.step_date)
yy_list = ptime.yyyymmdd2years(date_list)
flag_array = np.array(yy_list) >= step_yy
A_step = np.zeros((date_num, 1))
A_step[flag_array] = 1.0
A_def = np.hstack((A_def, A_step))
# Heresh's original code for phase history approach
#A_def = np.hstack((A2,A1,np.ones((date_num,1))))
print '-------------------------------------------------'
##---------------------------------------- Loop for L2-norm inversion -----------------------------------##
delta_z_mat = np.zeros([length, width], dtype=np.float32)
resid_n = np.zeros([A_def.shape[0], length * width], dtype=np.float32)
constC = np.zeros([length, width], dtype=np.float32)
#delta_a_mat = np.zeros([length, width])
if inps.incidence_angle.ndim == 2 and inps.range_dis.ndim == 2:
print 'inversing using L2-norm minimization (unweighted least squares)'\
' pixel by pixel: %d loops in total' % (length*width)
prog_bar = ptime.progress_bar(maxValue=length * width,
prefix='calculating: ')
for i in range(length * width):
row = i % length
col = i / length
range_dis = inps.range_dis[row, col]
inc_angle = inps.incidence_angle[row, col]
# Consider P_BASELINE variation within one interferogram
if inps.pbase.shape[1] > 1:
pbase = inps.pbase[:, row].reshape(date_num, 1)
# Design matrix - DEM error using pbase, range distance and incidence angle
A_delta_z = pbase / (range_dis * np.sin(inc_angle))
if inps.phase_velocity:
pbase_v = np.diff(pbase, axis=0) / np.diff(inps.tbase, axis=0)
A_delta_z_v = pbase_v / (range_dis * np.sin(inc_angle))
A = np.hstack((A_delta_z_v, A_def))
else:
A = np.hstack((A_delta_z, A_def))
# L-2 norm inversion
if inps.ex_date:
A_inv = np.linalg.pinv(A[inps.ex_flag, :])
else:
A_inv = np.linalg.pinv(A)
# Get unknown parameters X = [delta_z, vel, acc, delta_acc, ...]
ts_dis = timeseries[:, i]
if inps.phase_velocity:
ts_dis = np.diff(ts_dis, axis=0) / np.diff(inps.tbase, axis=0)
if inps.ex_date:
X = np.dot(A_inv, ts_dis[inps.ex_flag])
else:
X = np.dot(A_inv, ts_dis)
# Residual vector n
resid_n[:, i] = ts_dis - np.dot(A, X)
# Update DEM error / timeseries matrix
delta_z = X[0]
delta_z_mat[row, col] = delta_z
if inps.update_timeseries:
timeseries[:, i] -= np.dot(A_delta_z, delta_z).flatten()
prog_bar.update(i + 1, every=length * width / 100)
prog_bar.close()
elif inps.incidence_angle.ndim == 1 and inps.range_dis.ndim == 1:
print 'inversing using L2-norm minimization (unweighted least squares)'\
' column by column: %d loops in total' % (width)
prog_bar = ptime.progress_bar(maxValue=width, prefix='calculating: ')
for i in range(width):
range_dis = inps.range_dis[i]
inc_angle = inps.incidence_angle[i]
# Design matrix - DEM error using pbase, range distance and incidence angle
A_delta_z = pbase / (range_dis * np.sin(inc_angle))
if inps.phase_velocity:
pbase_v = np.diff(pbase, axis=0) / np.diff(inps.tbase, axis=0)
A_delta_z_v = pbase_v / (range_dis * np.sin(inc_angle))
A = np.hstack((A_delta_z_v, A_def))
else:
A = np.hstack((A_delta_z, A_def))
# L-2 norm inversion
if inps.ex_date:
A_inv = np.linalg.pinv(A[inps.ex_flag, :])
else:
A_inv = np.linalg.pinv(A)
# Get unknown parameters X = [delta_z, vel, acc, delta_acc, ...]
ts_dis = timeseries[:, i * length:(i + 1) * length]
if inps.phase_velocity:
ts_dis = np.diff(ts_dis, axis=0) / np.diff(inps.tbase, axis=0)
if inps.ex_date:
X = np.dot(A_inv, ts_dis[inps.ex_flag, :])
else:
X = np.dot(A_inv, ts_dis)
# Residual vector n
resid_n[:, i * length:(i + 1) * length] = ts_dis - np.dot(A, X)
constC[:, i] = X[1].reshape((1, length))
# Update DEM error / timeseries matrix
delta_z = X[0].reshape((1, length))
delta_z_mat[:, i] = delta_z
if inps.update_timeseries:
timeseries[:, i * length:(i + 1) * length] -= np.dot(
A_delta_z, delta_z)
prog_bar.update(i + 1, every=width / 100)
prog_bar.close()
elif inps.incidence_angle.ndim == 0 and inps.range_dis.ndim == 0:
print 'inversing using L2-norm minimization (unweighted least squares) for the whole area'
# Design matrix - DEM error using pbase, range distance and incidence angle
A_delta_z = pbase / (inps.range_dis * np.sin(inps.incidence_angle))
if inps.phase_velocity:
pbase_v = np.diff(pbase, axis=0) / np.diff(inps.tbase, axis=0)
A_delta_z_v = pbase_v / (inps.range_dis *
np.sin(inps.incidence_angle))
A = np.hstack((A_delta_z_v, A_def))
else:
A = np.hstack((A_delta_z, A_def))
# L-2 norm inversion
if inps.ex_date:
A_inv = np.linalg.pinv(A[inps.ex_flag, :])
else:
A_inv = np.linalg.pinv(A)
# Get unknown parameters X = [delta_z, vel, acc, delta_acc, ...]
if inps.phase_velocity:
timeseries = np.diff(timeseries, axis=0) / np.diff(inps.tbase,
axis=0)
if inps.ex_date:
X = np.dot(A_inv, timeseries[inps.ex_flag, :])
else:
X = np.dot(A_inv, timeseries)
# Residual vector n
resid_n = ts_dis - np.dot(A, X)
# Update DEM error / timeseries matrix
delta_z_mat = X[0].reshape((1, length * width))
if inps.update_timeseries:
timeseries -= np.dot(A_delta_z, delta_z_mat)
delta_z_mat = np.reshape(delta_z_mat, [length, width], order='F')
else:
print 'ERROR: Script only support same dimension for both incidence angle and range distance matrix.'
print 'dimension of incidence angle: ' + str(inps.incidence_angle.ndim)
print 'dimension of range distance: ' + str(inps.range_dis.ndim)
sys.exit(1)
##------------------------------------------------ Output --------------------------------------------##
# DEM error file
if 'Y_FIRST' in atr.keys():
dem_error_file = 'demGeo_error.h5'
else:
dem_error_file = 'demRadar_error.h5'
#if inps.phase_velocity: suffix = '_pha_poly'+str(inps.poly_order)
#else: suffix = '_vel_poly'+str(inps.poly_order)
#dem_error_file = os.path.splitext(dem_error_file)[0]+suffix+os.path.splitext(dem_error_file)[1]
print 'writing >>> ' + dem_error_file
atr_dem_error = atr.copy()
atr_dem_error['FILE_TYPE'] = 'dem'
atr_dem_error['UNIT'] = 'm'
writefile.write(delta_z_mat, atr_dem_error, dem_error_file)
## Phase Constant C = resid_n[0,:]
#atrC = atr.copy()
#atrC['FILE_TYPE'] = 'mask'
#atrC['UNIT'] = 'm'
#writefile.write(constC, atrC, 'constD.h5')
## Corrected DEM file
#if inps.dem_file:
# inps.dem_outfile = os.path.splitext(inps.dem_file)[0]+suffix+os.path.splitext(inps.dem_file)[1]
# print '--------------------------------------'
# print 'writing >>> '+inps.dem_outfile
# dem, atr_dem = readfile.read(inps.dem_file)
# writefile.write(dem+delta_z_mat, atr_dem, inps.dem_outfile)
#outfile = 'delta_acc.h5'
#print 'writing >>> '+outfile
#atr_dem_error = atr.copy()
#atr_dem_error['FILE_TYPE'] = 'velocity'
#atr_dem_error['UNIT'] = 'm/s'
#writefile.write(delta_a_mat, atr_dem_error, outfile)
#print '**************************************'
# Corrected Time Series
if inps.update_timeseries:
print 'writing >>> ' + inps.outfile
print 'number of dates: ' + str(len(date_list))
h5out = h5py.File(inps.outfile, 'w')
group = h5out.create_group('timeseries')
prog_bar = ptime.progress_bar(maxValue=date_num, prefix='writing: ')
for i in range(date_num):
date = date_list[i]
d = np.reshape(timeseries[i][:], [length, width], order='F')
dset = group.create_dataset(date, data=d, compression='gzip')
prog_bar.update(i + 1, suffix=date)
prog_bar.close()
for key, value in atr.iteritems():
group.attrs[key] = value
h5out.close()
outFile = os.path.splitext(inps.outfile)[0] + 'InvResid.h5'
print 'writing >>> ' + outFile
print 'number of dates: ' + str(A_def.shape[0])
h5out = h5py.File(outFile, 'w')
group = h5out.create_group('timeseries')
prog_bar = ptime.progress_bar(maxValue=A_def.shape[0], prefix='writing: ')
for i in range(A_def.shape[0]):
date = date_list[i]
d = np.reshape(resid_n[i][:], [length, width], order='F')
dset = group.create_dataset(date, data=d, compression='gzip')
prog_bar.update(i + 1, suffix=date)
prog_bar.close()
# Attribute
for key, value in atr.iteritems():
group.attrs[key] = value
if A_def.shape[0] == date_num:
group.attrs['UNIT'] = 'm'
else:
group.attrs['UNIT'] = 'm/yr'
h5out.close()
return
def main(argv):
inps = cmdLineParse()
suffix = '_demErr'
if not inps.outfile:
inps.outfile = os.path.splitext(inps.timeseries_file)[0]+suffix+os.path.splitext(inps.timeseries_file)[1]
if inps.template_file:
print 'read option from template file: '+inps.template_file
inps = read_template2inps(inps.template_file, inps)
##### Read Data
atr = readfile.read_attribute(inps.timeseries_file)
coordType = 'radar'
if 'Y_FIRST' in atr.keys():
coordType = 'geo'
# 1. Incidence angle
try:
inps.inc_angle_file = ut.get_file_list(inps.inc_angle_file, coord=coordType)[0]
except ValueError:
print 'No incidence angle file found!\nRun incidence_angle.py to generate it.'
print 'read incidence angle from file: '+str(inps.inc_angle_file)
inps.inc_angle = readfile.read(inps.inc_angle_file, epoch='incidenceAngle')[0].flatten()
inps.inc_angle *= np.pi/180.0
# 2. Slant Range distance
try:
inps.range_dist_file = ut.get_file_list(inps.range_dist_file, coord=coordType)[0]
except ValueError:
print 'No range distance file found!\nRun range_distance.py to generate it.'
print 'read slant range distance from file: '+str(inps.range_dist_file)
inps.range_dist = readfile.read(inps.range_dist_file, epoch='slantRangeDistance')[0].flatten()
# 3. Perp Baseline - 1D in time, 0D/1D in space (azimuth)
print 'read perpendicular baseline'
try:
inps.pbase = ut.perp_baseline_timeseries(atr, dimension=1)
if inps.pbase.shape[1] > 1:
print 'consider perp baseline variance in azimuth direction'
except valueError:
print 'No P_BASELINE_TIMESERIES found in timeseries file.\n'+\
'Can not correct for DEM residula without it!'
# 4. Time Series - 1D in time, 1D in space (flattened)
print "read time series file: " + inps.timeseries_file
h5 = h5py.File(inps.timeseries_file)
date_list = sorted(h5['timeseries'].keys())
date_num = len(date_list)
inps.tbase = np.array(ptime.date_list2tbase(date_list)[0]).reshape(-1,1)
#Mark dates used in the estimation
inps.ex_date = check_exclude_date(inps.ex_date, date_list)
inps.date_flag = np.array([i not in inps.ex_date for i in date_list], dtype=np.bool_)
if inps.poly_order > np.sum(inps.date_flag):
raise ValueError("ERROR: input polynomial order=%d is larger than number of acquisition=%d used in estimation!" %\
(inps.poly_order, np.sum(inps.date_flag)))
length = int(atr['FILE_LENGTH'])
width = int(atr['WIDTH'])
pixel_num = length*width
timeseries = np.zeros((date_num, pixel_num),np.float32)
for i in range(date_num):
timeseries[i] = h5['timeseries'].get(date_list[i])[:].flatten()
sys.stdout.write('\rreading acquisition %3d/%3d ...' % (i+1, date_num))
sys.stdout.flush()
h5.close()
print ''
##### Design matrix - temporal deformation model
print '-------------------------------------------------'
print 'Correct topographic phase residual using Fattahi and Amelung (2013, IEEE-TGRS)'
msg = 'minimum-norm constrain on: phase'
if inps.phase_velocity:
msg += ' velocity'
print msg
# Heresh's original code for phase history approach
#A1 = np.hstack((np.ones((date_num, 1)), inps.tbase))
#A2 = inps.tbase**2 / 2.0
#A_def = np.hstack((A2,A1,np.ones((date_num,1))))
# 1. Polynomial - 2D matrix in size of (date_num, polyOrder+1)
print "temporal deformation model: polynomial order = "+str(inps.poly_order)
A_def = np.ones((date_num, 1), np.float32)
for i in range(inps.poly_order):
Ai = inps.tbase**(i+1) / gamma(i+2)
Ai = np.array(Ai, np.float32).reshape(-1,1)
A_def = np.hstack((A_def, Ai))
# 2. Step function - 2D matrix in size of (date_num, stepNum)
if inps.step_date:
print "temporal deformation model: step functions at "+str(inps.step_date)
yySteps = ptime.yyyymmdd2years(inps.step_date)
yyList = np.array(ptime.yyyymmdd2years(date_list)).reshape(-1,1)
for yyStep in yySteps:
Ai = yyList > yyStep
Ai = np.array(Ai, np.float32).reshape(-1,1)
A_def = np.hstack((A_def, Ai))
inps.step_num = len(inps.step_date)
print '-------------------------------------------------'
##---------------------------------------- Loop for L2-norm inversion -----------------------------------##
## Output estimated steps
print 'ordinal least squares (OLS) inversion using L2-norm minimization'
timeseriesCor = np.zeros((date_num, pixel_num), dtype=np.float32)
timeseriesRes = np.zeros((date_num, pixel_num), dtype=np.float32)
topoRes = np.zeros(pixel_num, dtype=np.float32)
constC = np.zeros(pixel_num, dtype=np.float32)
if inps.step_num > 0:
stepModel = np.zeros((inps.step_num, pixel_num), dtype=np.float32)
print 'skip pixels with zero/nan value in geometry files - incidence angle and range distance'
mask = np.multiply(~np.isnan(inps.inc_angle), ~np.isnan(inps.range_dist))
mask[inps.inc_angle == 0.] = 0
mask[inps.range_dist == 0.] = 0
pixel_num2inv = np.sum(mask)
pixel_idx2inv = np.where(mask)[0]
print 'number of pixels in the file: %d' % (pixel_num)
print 'number of pixels to inverse: %d' % (pixel_num2inv)
if inps.pbase.shape[1] == 1:
pbase = inps.pbase
prog_bar = ptime.progress_bar(maxValue=pixel_num)
for i in range(pixel_num2inv):
prog_bar.update(i+1, every=1000, suffix='%s/%s pixels'%(str(i+1), str(pixel_num2inv)))
idx = pixel_idx2inv[i]
r = inps.range_dist[idx]
inc_angle = inps.inc_angle[idx]
if inps.pbase.shape[1] > 1:
pbase = inps.pbase[:, int(idx/width)].reshape(-1,1)
A_deltaZ = pbase / (r * np.sin(inc_angle))
A = np.hstack((A_deltaZ, A_def))
ts = timeseries[:,idx].reshape(date_num,-1)
deltaZ, tsCor, tsRes, stepEst = topographic_residual_inversion(ts, A, inps)
topoRes[idx:idx+1] = deltaZ
timeseriesCor[:,idx:idx+1] = tsCor
timeseriesRes[:,idx:idx+1] = tsRes
if inps.step_num > 0:
stepModel[:,idx:idx+1] = stepEst
prog_bar.close()
##------------------------------------------------ Output --------------------------------------------##
# 1. DEM error file
if 'Y_FIRST' in atr.keys():
deltaZFile = 'demGeo_error.h5'
else:
deltaZFile = 'demRadar_error.h5'
print 'writing >>> '+deltaZFile
atrDeltaZ = atr.copy()
atrDeltaZ['FILE_TYPE'] = 'dem'
atrDeltaZ['UNIT'] = 'm'
writefile.write(topoRes.reshape(length, width), atrDeltaZ, deltaZFile)
# 2. Topo Residual Corrected Time Series
print 'writing >>> '+inps.outfile
h5 = h5py.File(inps.outfile,'w')
group = h5.create_group('timeseries')
for i in range(date_num):
sys.stdout.write('\rwriting acquisition %3d/%3d ...' % (i+1, date_num))
sys.stdout.flush()
dset = group.create_dataset(date_list[i], data=timeseriesCor[i].reshape(length, width), compression='gzip')
print ''
for key,value in atr.iteritems():
group.attrs[key] = value
h5.close()
# 3. Inversion residual Time Series
tsResFile = os.path.join(os.path.dirname(inps.outfile), 'timeseriesResidual.h5')
print 'writing >>> '+os.path.basename(tsResFile)
h5 = h5py.File(tsResFile,'w')
group = h5.create_group('timeseries')
for i in range(date_num):
sys.stdout.write('\rwriting acquisition %3d/%3d ...' % (i+1, date_num))
sys.stdout.flush()
dset = group.create_dataset(date_list[i], data=timeseriesRes[i].reshape(length, width), compression='gzip')
print ''
# Attribute
for key,value in atr.iteritems():
group.attrs[key] = value
h5.close()
# 4. Step temporal Model estimation
if inps.step_num > 0:
stepFile = os.path.join(os.path.dirname(inps.outfile), 'timeseriesStepModel.h5')
print 'writing >>> '+os.path.basename(stepFile)
h5 = h5py.File(stepFile,'w')
group = h5.create_group('timeseries')
for i in range(inps.step_num):
sys.stdout.write('\rwriting acquisition %3d/%3d ...' % (i+1, inps.step_num))
sys.stdout.flush()
dset = group.create_dataset(inps.step_date[i], data=stepModel[i].reshape(length, width), compression='gzip')
print ''
# Attribute
for key,value in atr.iteritems():
group.attrs[key] = value
group.attrs.pop('ref_date')
h5.close()
print 'Done.'
return