def read_baseline_info(baseline_file, reference_file):
"""Read date, bperp and/or DOP info
Parameters: baseline_file : str, path of bl_list.txt file
reference_file : str, path of ifgramStack.h5 file
Returns: date_list : list of str in YYMMDD format
tbase_list : list of int in days
pbase_list : list of float in meter
dop_list : None, list of 1D array in size of (3,)
"""
dop_list = None
if baseline_file:
date_list, pbase_list, dop_list = pnet.read_baseline_file(
baseline_file)[0:3]
date_list = ptime.yymmdd(date_list)
tbase_list = ptime.date_list2tbase(date_list)[0]
elif reference_file:
obj = ifgramStack(reference_file)
date12_list_all = obj.get_date12_list(dropIfgram=False)
date12_list_all = ptime.yymmdd_date12(date12_list_all)
m_dates = [i.split('-')[0] for i in date12_list_all]
s_dates = [i.split('-')[1] for i in date12_list_all]
date_list = sorted(list(set(m_dates + s_dates)))
tbase_list = ptime.date_list2tbase(date_list)[0]
pbase_list = obj.get_perp_baseline_timeseries(
dropIfgram=False).tolist()
return date_list, tbase_list, pbase_list, dop_list
def select_pairs_delaunay(date_list, pbase_list, norm=True, date12_format='YYMMDD-YYMMDD'):
"""Select Pairs using Delaunay Triangulation based on temporal/perpendicular baselines
Inputs:
date_list : list of date in YYMMDD/YYYYMMDD format
pbase_list : list of float, perpendicular spatial baseline
norm : normalize temporal baseline to perpendicular baseline
Key points
1. Define a ratio between perpendicular and temporal baseline axis units (Pepe and Lanari, 2006, TGRS).
2. Pairs with too large perpendicular / temporal baseline or Doppler centroid difference should be removed
after this, using a threshold, to avoid strong decorrelations (Zebker and Villasenor, 1992, TGRS).
Reference:
Pepe, A., and R. Lanari (2006), On the extension of the minimum cost flow algorithm for phase unwrapping
of multitemporal differential SAR interferograms, IEEE TGRS, 44(9), 2374-2383.
Zebker, H. A., and J. Villasenor (1992), Decorrelation in interferometric radar echoes, IEEE TGRS, 30(5), 950-959.
"""
# Get temporal baseline in days
date6_list = ptime.yymmdd(date_list)
date8_list = ptime.yyyymmdd(date_list)
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Normalization (Pepe and Lanari, 2006, TGRS)
if norm:
temp2perp_scale = (max(pbase_list)-min(pbase_list)) / (max(tbase_list)-min(tbase_list))
tbase_list = [tbase*temp2perp_scale for tbase in tbase_list]
# Generate Delaunay Triangulation
date12_idx_list = Triangulation(tbase_list, pbase_list).edges.tolist()
date12_idx_list = [sorted(idx) for idx in sorted(date12_idx_list)]
# Convert index into date12
date12_list = [date6_list[idx[0]]+'-'+date6_list[idx[1]]
for idx in date12_idx_list]
if date12_format == 'YYYYMMDD_YYYYMMDD':
date12_list = ptime.yyyymmdd_date12(date12_list)
return date12_list
def velocity2timeseries(date_list, vel=0.03, display=False):
'''Simulate displacement time-series from linear velocity
Inputs:
date_list - list of string in YYYYMMDD or YYMMDD format
vel - float, velocity in meter per year
display - bool, display simulation or not
Output:
ts - 2D np.array in size of (date_num,1), displacement time-series in m
Example:
date_list = pnet.read_baseline_file('bl_list.txt')[0]
ts0 = simulate_timeseries(date_list, vel=0.03, display=True)
'''
tbase_list = ptime.date_list2tbase(date_list)[0]
ts = vel / 365.25 * np.array(tbase_list)
ts = ts.reshape(-1,1)
if display:
dates = ptime.date_list2vector(date_list)[0]
## Display
marker_size = 5
plt.figure()
plt.scatter(dates, ts*100.0, s=marker_size**2)
plt.xlabel('Time (years)')
plt.ylabel('LOS Displacement (cm)')
plt.title('Displacement time-series with velocity = '+str(vel)+' m/yr')
plt.show()
return ts
def velocity2timeseries(date_list, vel=0.03, display=False):
'''Simulate displacement time-series from linear velocity
Inputs:
date_list - list of string in YYYYMMDD or YYMMDD format
vel - float, velocity in meter per year
display - bool, display simulation or not
Output:
ts - 2D np.array in size of (date_num,1), displacement time-series in m
Example:
date_list = pnet.read_baseline_file('bl_list.txt')[0]
ts0 = simulate_timeseries(date_list, vel=0.03, display=True)
'''
tbase_list = ptime.date_list2tbase(date_list)[0]
ts = vel / 365.25 * np.array(tbase_list)
ts = ts.reshape(-1, 1)
if display:
dates = ptime.date_list2vector(date_list)[0]
## Display
marker_size = 5
plt.figure()
plt.scatter(dates, ts * 100.0, s=marker_size**2)
plt.xlabel('time [years]')
plt.ylabel('LOS displacement [cm]')
plt.title('displacement time-series with velocity = ' + str(vel) +
' m/yr')
plt.show()
return ts
def select_master_date(date_list, pbase_list=[]):
"""Select super master date based on input temporal and/or perpendicular baseline info.
Return master date in YYYYMMDD format.
"""
date8_list = ptime.yyyymmdd(date_list)
if not pbase_list:
# Choose date in the middle
m_date8 = date8_list[int(len(date8_list)/2)]
else:
# Get temporal baseline list
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Normalization (Pepe and Lanari, 2006, TGRS)
temp2perp_scale = (max(pbase_list)-min(pbase_list)) / (max(tbase_list)-min(tbase_list))
tbase_list = [tbase*temp2perp_scale for tbase in tbase_list]
# Get distance matrix
ttMat1, ttMat2 = np.meshgrid(np.array(tbase_list),
np.array(tbase_list))
ppMat1, ppMat2 = np.meshgrid(np.array(pbase_list),
np.array(pbase_list))
ttMat = np.abs(ttMat1 - ttMat2) # temporal distance matrix
ppMat = np.abs(ppMat1 - ppMat2) # spatial distance matrix
# 2D distance matrix in temp/perp domain
disMat = np.sqrt(np.square(ttMat) + np.square(ppMat))
# Choose date minimize the total distance of temp/perp baseline
disMean = np.mean(disMat, 0)
m_idx = np.argmin(disMean)
m_date8 = date8_list[m_idx]
return m_date8
def select_pairs_delaunay(date_list, pbase_list, norm=True, date12_format='YYMMDD-YYMMDD'):
"""Select Pairs using Delaunay Triangulation based on temporal/perpendicular baselines
Inputs:
date_list : list of date in YYMMDD/YYYYMMDD format
pbase_list : list of float, perpendicular spatial baseline
norm : normalize temporal baseline to perpendicular baseline
Key points
1. Define a ratio between perpendicular and temporal baseline axis units (Pepe and Lanari, 2006, TGRS).
2. Pairs with too large perpendicular / temporal baseline or Doppler centroid difference should be removed
after this, using a threshold, to avoid strong decorrelations (Zebker and Villasenor, 1992, TGRS).
Reference:
Pepe, A., and R. Lanari (2006), On the extension of the minimum cost flow algorithm for phase unwrapping
of multitemporal differential SAR interferograms, IEEE TGRS, 44(9), 2374-2383.
Zebker, H. A., and J. Villasenor (1992), Decorrelation in interferometric radar echoes, IEEE TGRS, 30(5), 950-959.
"""
# Get temporal baseline in days
date6_list = ptime.yymmdd(date_list)
date8_list = ptime.yyyymmdd(date_list)
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Normalization (Pepe and Lanari, 2006, TGRS)
if norm:
temp2perp_scale = (max(pbase_list)-min(pbase_list)) / (max(tbase_list)-min(tbase_list))
tbase_list = [tbase*temp2perp_scale for tbase in tbase_list]
# Generate Delaunay Triangulation
date12_idx_list = Triangulation(tbase_list, pbase_list).edges.tolist()
date12_idx_list = [sorted(idx) for idx in sorted(date12_idx_list)]
# Convert index into date12
date12_list = [date6_list[idx[0]]+'-'+date6_list[idx[1]]
for idx in date12_idx_list]
if date12_format == 'YYYYMMDD_YYYYMMDD':
date12_list = ptime.yyyymmdd_date12(date12_list)
return date12_list
def select_master_date(date_list, pbase_list=[]):
"""Select super master date based on input temporal and/or perpendicular baseline info.
Return master date in YYYYMMDD format.
"""
date8_list = ptime.yyyymmdd(date_list)
if not pbase_list:
# Choose date in the middle
m_date8 = date8_list[int(len(date8_list) / 2)]
else:
# Get temporal baseline list
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Normalization (Pepe and Lanari, 2006, TGRS)
temp2perp_scale = (max(pbase_list) - min(pbase_list)) / (
max(tbase_list) - min(tbase_list))
tbase_list = [tbase * temp2perp_scale for tbase in tbase_list]
# Get distance matrix
ttMat1, ttMat2 = np.meshgrid(np.array(tbase_list),
np.array(tbase_list))
ppMat1, ppMat2 = np.meshgrid(np.array(pbase_list),
np.array(pbase_list))
ttMat = np.abs(ttMat1 - ttMat2) # temporal distance matrix
ppMat = np.abs(ppMat1 - ppMat2) # spatial distance matrix
# 2D distance matrix in temp/perp domain
disMat = np.sqrt(np.square(ttMat) + np.square(ppMat))
# Choose date minimize the total distance of temp/perp baseline
disMean = np.mean(disMat, 0)
m_idx = np.argmin(disMean)
m_date8 = date8_list[m_idx]
return m_date8
def threshold_temporal_baseline(date12_list, btemp_max, keep_seasonal=True, btemp_min=0.0):
"""Remove pairs/interferograms out of min/max/seasonal temporal baseline limits
Inputs:
date12_list : list of string for date12 in YYMMDD-YYMMDD format
btemp_max : float, maximum temporal baseline
btemp_min : float, minimum temporal baseline
keep_seasonal : keep interferograms with seasonal temporal baseline
Output:
date12_list_out : list of string for date12 in YYMMDD-YYMMDD format
Example:
date12_list = threshold_temporal_baseline(date12_list, 200)
date12_list = threshold_temporal_baseline(date12_list, 200, False)
"""
if not date12_list:
return []
# Get date list and tbase list
m_dates = [date12.split('-')[0] for date12 in date12_list]
s_dates = [date12.split('-')[1] for date12 in date12_list]
date8_list = sorted(ptime.yyyymmdd(list(set(m_dates + s_dates))))
date6_list = ptime.yymmdd(date8_list)
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Threshold
date12_list_out = []
for date12 in date12_list:
date1, date2 = date12.split('-')
idx1 = date6_list.index(date1)
idx2 = date6_list.index(date2)
tbase = int(abs(tbase_list[idx1] - tbase_list[idx2]))
if btemp_min <= tbase <= btemp_max:
date12_list_out.append(date12)
elif keep_seasonal and tbase/30 in [11, 12]:
date12_list_out.append(date12)
return date12_list_out
def threshold_temporal_baseline(date12_list, btemp_max, keep_seasonal=True, btemp_min=0.0):
"""Remove pairs/interferograms out of min/max/seasonal temporal baseline limits
Inputs:
date12_list : list of string for date12 in YYMMDD-YYMMDD format
btemp_max : float, maximum temporal baseline
btemp_min : float, minimum temporal baseline
keep_seasonal : keep interferograms with seasonal temporal baseline
Output:
date12_list_out : list of string for date12 in YYMMDD-YYMMDD format
Example:
date12_list = threshold_temporal_baseline(date12_list, 200)
date12_list = threshold_temporal_baseline(date12_list, 200, False)
"""
if not date12_list:
return []
# Get date list and tbase list
m_dates = [date12.split('-')[0] for date12 in date12_list]
s_dates = [date12.split('-')[1] for date12 in date12_list]
date8_list = sorted(ptime.yyyymmdd(list(set(m_dates + s_dates))))
date6_list = ptime.yymmdd(date8_list)
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Threshold
date12_list_out = []
for date12 in date12_list:
date1, date2 = date12.split('-')
idx1 = date6_list.index(date1)
idx2 = date6_list.index(date2)
tbase = int(abs(tbase_list[idx1] - tbase_list[idx2]))
if btemp_min <= tbase <= btemp_max:
date12_list_out.append(date12)
elif keep_seasonal and tbase/30 in [11, 12]:
date12_list_out.append(date12)
return date12_list_out
def simulate_coherence_v2(date12_list,
decor_time=200.0,
coh_resid=0.2,
inc_angle=40,
sensor_name='Sen',
display=False):
"""Simulate coherence version 2 (without using bl_list.txt file).
Parameters: date12_list - list of string in YYYYMMDD_YYYYMMDD format, indicating pairs configuration
decor_time - float, decorrelation rate in days, time for coherence to drop to 1/e of its initial value
coh_resid - float, long-term coherence, minimum attainable coherence value
inc_angle - float, incidence angle in degrees
sensor_name - string, SAR sensor name
display - bool, display result as matrix or not
Returns: coh - 2D np.array in size of (ifgram_num)
"""
num_pair = len(date12_list)
date1s = [x.split('_')[0] for x in date12_list]
date2s = [x.split('_')[1] for x in date12_list]
date_list = sorted(list(set(date1s + date2s)))
tbase_list = ptime.date_list2tbase(date_list)[0]
SNR = 22 # NESZ = -22 dB from Table 1 in https://sentinels.copernicus.eu/web/sentinel/
coh_thermal = 1. / (1. + 1. / SNR)
# bperp
rng = np.random.default_rng(2)
pbase_list = rng.normal(0, 50, num_pair).tolist()
pbase_c = critical_perp_baseline(sensor_name, inc_angle)
coh = np.zeros(num_pair, dtype=np.float32)
for i in range(num_pair):
date1, date2 = date12_list[i].split('_')
ind1, ind2 = date_list.index(date1), date_list.index(date2)
tbase = tbase_list[ind2] - tbase_list[ind1]
pbase = pbase_list[ind2] - pbase_list[ind1]
coh_geom = (pbase_c - abs(pbase)) / pbase_c
coh_temp = np.multiply(
(coh_thermal - coh_resid), np.exp(
-1 * abs(tbase) / decor_time)) + coh_resid
coh[i] = coh_geom * coh_temp
if display:
print(('critical perp baseline: %.f m' % pbase_c))
coh_mat = coherence_matrix(date12_list, coh)
plt.figure()
plt.imshow(coh_mat, vmin=0.0, vmax=1.0, cmap='jet')
plt.xlabel('Image number')
plt.ylabel('Image number')
cbar = plt.colorbar()
cbar.set_label('Coherence')
plt.title('Coherence matrix')
plt.show()
return coh
def select_master_interferogram(date12_list,
date_list,
pbase_list,
m_date=None):
"""Select reference interferogram based on input temp/perp baseline info
If master_date is specified, select its closest slave_date, which is newer than master_date;
otherwise, choose the closest pair among all pairs as master interferogram.
Example:
master_date12 = pnet.select_master_ifgram(date12_list, date_list, pbase_list)
'080211-080326' = pnet.select_master_ifgram(date12_list, date_list, pbase_list, m_date='080211')
"""
pbase_array = np.array(pbase_list, dtype='float64')
# Get temporal baseline
date8_list = ptime.yyyymmdd(date_list)
date6_list = ptime.yymmdd(date8_list)
tbase_array = np.array(ptime.date_list2tbase(date8_list)[0],
dtype='float64')
# Normalization (Pepe and Lanari, 2006, TGRS)
temp2perp_scale = (max(pbase_array) - min(pbase_array)) / (
max(tbase_array) - min(tbase_array))
tbase_array *= temp2perp_scale
# Calculate sqrt of temp/perp baseline for input pairs
idx1 = np.array(
[date6_list.index(date12.split('-')[0]) for date12 in date12_list])
idx2 = np.array(
[date6_list.index(date12.split('-')[1]) for date12 in date12_list])
base_distance = np.sqrt((tbase_array[idx2] - tbase_array[idx1])**2 +
(pbase_array[idx2] - pbase_array[idx1])**2)
# Get master interferogram index
if not m_date:
# Choose pair with shortest temp/perp baseline
m_date12_idx = np.argmin(base_distance)
else:
m_date = ptime.yymmdd(m_date)
# Choose pair contains m_date with shortest temp/perp baseline
m_date12_idx_array = np.array([
date12_list.index(date12) for date12 in date12_list
if m_date + '-' in date12
])
min_base_distance = np.min(base_distance[m_date12_idx_array])
m_date12_idx = np.where(base_distance == min_base_distance)[0][0]
m_date12 = date12_list[m_date12_idx]
return m_date12
def select_pairs_mst(date_list, pbase_list, date12_format='YYMMDD-YYMMDD'):
"""Select Pairs using Minimum Spanning Tree technique
Connection Cost is calculated using the baseline distance in perp and scaled temporal baseline (Pepe and Lanari,
2006, TGRS) plane.
Inputs:
date_list : list of date in YYMMDD/YYYYMMDD format
pbase_list : list of float, perpendicular spatial baseline
References:
Pepe, A., and R. Lanari (2006), On the extension of the minimum cost flow algorithm for phase unwrapping
of multitemporal differential SAR interferograms, IEEE TGRS, 44(9), 2374-2383.
Perissin D., Wang T. (2012), Repeat-pass SAR interferometry with partially coherent targets. IEEE TGRS. 271-280
"""
# Get temporal baseline in days
date6_list = ptime.yymmdd(date_list)
date8_list = ptime.yyyymmdd(date_list)
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Normalization (Pepe and Lanari, 2006, TGRS)
temp2perp_scale = (max(pbase_list) - min(pbase_list)) / (max(tbase_list) -
min(tbase_list))
tbase_list = [tbase * temp2perp_scale for tbase in tbase_list]
# Get weight matrix
ttMat1, ttMat2 = np.meshgrid(np.array(tbase_list), np.array(tbase_list))
ppMat1, ppMat2 = np.meshgrid(np.array(pbase_list), np.array(pbase_list))
ttMat = np.abs(ttMat1 - ttMat2) # temporal distance matrix
ppMat = np.abs(ppMat1 - ppMat2) # spatial distance matrix
# 2D distance matrix in temp/perp domain
weightMat = np.sqrt(np.square(ttMat) + np.square(ppMat))
weightMat = sparse.csr_matrix(weightMat) # compress sparse row matrix
# MST path based on weight matrix
mstMat = sparse.csgraph.minimum_spanning_tree(weightMat)
# Convert MST index matrix into date12 list
[s_idx_list, m_idx_list] = [
date_idx_array.tolist() for date_idx_array in sparse.find(mstMat)[0:2]
]
date12_list = []
for i in range(len(m_idx_list)):
idx = sorted([m_idx_list[i], s_idx_list[i]])
date12 = date6_list[idx[0]] + '-' + date6_list[idx[1]]
date12_list.append(date12)
if date12_format == 'YYYYMMDD_YYYYMMDD':
date12_list = ptime.yyyymmdd_date12(date12_list)
return date12_list
def select_pairs_mst(date_list, pbase_list, date12_format='YYMMDD-YYMMDD'):
"""Select Pairs using Minimum Spanning Tree technique
Connection Cost is calculated using the baseline distance in perp and scaled temporal baseline (Pepe and Lanari,
2006, TGRS) plane.
Inputs:
date_list : list of date in YYMMDD/YYYYMMDD format
pbase_list : list of float, perpendicular spatial baseline
References:
Pepe, A., and R. Lanari (2006), On the extension of the minimum cost flow algorithm for phase unwrapping
of multitemporal differential SAR interferograms, IEEE TGRS, 44(9), 2374-2383.
Perissin D., Wang T. (2012), Repeat-pass SAR interferometry with partially coherent targets. IEEE TGRS. 271-280
"""
# Get temporal baseline in days
date6_list = ptime.yymmdd(date_list)
date8_list = ptime.yyyymmdd(date_list)
tbase_list = ptime.date_list2tbase(date8_list)[0]
# Normalization (Pepe and Lanari, 2006, TGRS)
temp2perp_scale = (max(pbase_list)-min(pbase_list)) / (max(tbase_list)-min(tbase_list))
tbase_list = [tbase*temp2perp_scale for tbase in tbase_list]
# Get weight matrix
ttMat1, ttMat2 = np.meshgrid(np.array(tbase_list), np.array(tbase_list))
ppMat1, ppMat2 = np.meshgrid(np.array(pbase_list), np.array(pbase_list))
ttMat = np.abs(ttMat1 - ttMat2) # temporal distance matrix
ppMat = np.abs(ppMat1 - ppMat2) # spatial distance matrix
# 2D distance matrix in temp/perp domain
weightMat = np.sqrt(np.square(ttMat) + np.square(ppMat))
weightMat = sparse.csr_matrix(weightMat) # compress sparse row matrix
# MST path based on weight matrix
mstMat = sparse.csgraph.minimum_spanning_tree(weightMat)
# Convert MST index matrix into date12 list
[s_idx_list, m_idx_list] = [date_idx_array.tolist()
for date_idx_array in sparse.find(mstMat)[0:2]]
date12_list = []
for i in range(len(m_idx_list)):
idx = sorted([m_idx_list[i], s_idx_list[i]])
date12 = date6_list[idx[0]]+'-'+date6_list[idx[1]]
date12_list.append(date12)
if date12_format == 'YYYYMMDD_YYYYMMDD':
date12_list = ptime.yyyymmdd_date12(date12_list)
return date12_list
def get_design_matrix4time_func(date_list, poly_order=2, step_func_dates=[]):
"""get design matrix for temporal deformation model"""
# temporal baseline in the unit of years
tbase = ptime.date_list2tbase(date_list)[0]
tbase = np.array(tbase, dtype=np.float32) / 365.25
# 1. Polynomial - 2D matrix in size of (numDate, polyOrder+1)
A_def = np.ones((len(date_list), 1), np.float32)
for i in range(poly_order):
Ai = np.array(tbase**(i + 1) / gamma(i + 2), np.float32).reshape(-1, 1)
A_def = np.hstack((A_def, Ai))
# 2. Step function - 2D matrix in size of (numDate, len(step_func_dates))
if step_func_dates:
t_steps = ptime.yyyymmdd2years(step_func_dates)
t = np.array(ptime.yyyymmdd2years(date_list))
for t_step in t_steps:
Ai = np.array(t > t_step, np.float32).reshape(-1, 1)
A_def = np.hstack((A_def, Ai))
return A_def
def select_master_interferogram(date12_list, date_list, pbase_list, m_date=None):
"""Select reference interferogram based on input temp/perp baseline info
If master_date is specified, select its closest slave_date, which is newer than master_date;
otherwise, choose the closest pair among all pairs as master interferogram.
Example:
master_date12 = pnet.select_master_ifgram(date12_list, date_list, pbase_list)
'080211-080326' = pnet.select_master_ifgram(date12_list, date_list, pbase_list, m_date='080211')
"""
pbase_array = np.array(pbase_list, dtype='float64')
# Get temporal baseline
date8_list = ptime.yyyymmdd(date_list)
date6_list = ptime.yymmdd(date8_list)
tbase_array = np.array(ptime.date_list2tbase(date8_list)[0], dtype='float64')
# Normalization (Pepe and Lanari, 2006, TGRS)
temp2perp_scale = (max(pbase_array)-min(pbase_array)) / (max(tbase_array)-min(tbase_array))
tbase_array *= temp2perp_scale
# Calculate sqrt of temp/perp baseline for input pairs
idx1 = np.array([date6_list.index(date12.split('-')[0]) for date12 in date12_list])
idx2 = np.array([date6_list.index(date12.split('-')[1]) for date12 in date12_list])
base_distance = np.sqrt((tbase_array[idx2] - tbase_array[idx1])**2 +
(pbase_array[idx2] - pbase_array[idx1])**2)
# Get master interferogram index
if not m_date:
# Choose pair with shortest temp/perp baseline
m_date12_idx = np.argmin(base_distance)
else:
m_date = ptime.yymmdd(m_date)
# Choose pair contains m_date with shortest temp/perp baseline
m_date12_idx_array = np.array([date12_list.index(date12) for date12 in date12_list
if m_date+'-' in date12])
min_base_distance = np.min(base_distance[m_date12_idx_array])
m_date12_idx = np.where(base_distance == min_base_distance)[0][0]
m_date12 = date12_list[m_date12_idx]
return m_date12
def ifgram_inversion_patch(ifgram_file,
box=None,
ref_phase=None,
obs_ds_name='unwrapPhase',
weight_func='var',
water_mask_file=None,
min_norm_velocity=True,
mask_ds_name=None,
mask_threshold=0.4,
min_redundancy=1.0):
"""Invert one patch of an ifgram stack into timeseries.
Parameters: box - tuple of 4 int, indicating (x0, y0, x1, y1) of the area of interest
or None for the whole image
ifgram_file - str, interferograms stack HDF5 file, e.g. ./inputs/ifgramStack.h5
ref_phase - 1D array in size of (num_ifgram), or None
obs_ds_name - str, dataset to feed the inversion.
weight_func - str, weight function, choose in ['no', 'fim', 'var', 'coh']
water_mask_file - str, water mask filename if available, to skip inversion on water
min_norm_velocity - bool, minimize the residual phase or phase velocity
mask_ds_name - str, dataset name in ifgram_file used to mask unwrapPhase pixelwisely
mask_threshold - float, min coherence of pixels if mask_dataset_name='coherence'
min_redundancy - float, the min number of ifgrams for every acquisition.
Returns: ts - 3D array in size of (num_date, num_row, num_col)
temp_coh - 2D array in size of (num_row, num_col)
num_inv_ifg - 2D array in size of (num_row, num_col)
box - tuple of 4 int
Example: ifgram_inversion_patch('ifgramStack.h5', box=(0,200,1316,400))
"""
stack_obj = ifgramStack(ifgram_file)
stack_obj.open(print_msg=False)
# debug
#y, x = 258, 454
#box = (x, y, x+1, y+1)
## 1. input info
# size
if box:
num_row = box[3] - box[1]
num_col = box[2] - box[0]
else:
num_row = stack_obj.length
num_col = stack_obj.width
num_pixel = num_row * num_col
# get tbase_diff
date_list = stack_obj.get_date_list(dropIfgram=True)
num_date = len(date_list)
tbase = np.array(ptime.date_list2tbase(date_list)[0], np.float32) / 365.25
tbase_diff = np.diff(tbase).reshape(-1, 1)
# design matrix
date12_list = stack_obj.get_date12_list(dropIfgram=True)
A, B = stack_obj.get_design_matrix4timeseries(date12_list=date12_list)[0:2]
# prep for decor std time-series
#if os.path.isfile('reference_date.txt'):
# ref_date = str(np.loadtxt('reference_date.txt', dtype=bytes).astype(str))
#else:
# ref_date = date_list[0]
#Astd = stack_obj.get_design_matrix4timeseries(date12_list=date12_list, refDate=ref_date)[0]
#ref_idx = date_list.index(ref_date)
#time_idx = [i for i in range(num_date)]
#time_idx.remove(ref_idx)
# skip zero value in the network inversion for phase
if 'phase' in obs_ds_name.lower():
skip_zero_value = True
else:
skip_zero_value = False
# 1.1 read / calcualte weight
if weight_func in ['no', 'sbas']:
weight = None
else:
weight = calc_weight(stack_obj,
box,
weight_func=weight_func,
dropIfgram=True,
chunk_size=100000)
# 1.2 read / mask unwrapPhase / offset
pha_data = read_unwrap_phase(stack_obj,
box,
ref_phase,
obs_ds_name=obs_ds_name,
dropIfgram=True)
pha_data = mask_unwrap_phase(pha_data,
stack_obj,
box,
dropIfgram=True,
mask_ds_name=mask_ds_name,
mask_threshold=mask_threshold)
# 1.3 mask of pixels to invert
mask = np.ones(num_pixel, np.bool_)
# 1.3.1 - Water Mask
if water_mask_file:
print('skip pixels on water with mask from file: {}'.format(
os.path.basename(water_mask_file)))
atr_msk = readfile.read_attribute(water_mask_file)
len_msk, wid_msk = int(atr_msk['LENGTH']), int(atr_msk['WIDTH'])
if (len_msk, wid_msk) != (stack_obj.length, stack_obj.width):
raise ValueError(
'Input water mask file has different size from ifgramStack file.'
)
dsName = [
i for i in readfile.get_dataset_list(water_mask_file)
if i in ['waterMask', 'mask']
][0]
waterMask = readfile.read(water_mask_file, datasetName=dsName,
box=box)[0].flatten()
mask *= np.array(waterMask, dtype=np.bool_)
del waterMask
# 1.3.2 - Mask for Zero Phase in ALL ifgrams
if 'phase' in obs_ds_name.lower():
print('skip pixels with zero/nan value in all interferograms')
with warnings.catch_warnings():
# ignore warning message for all-NaN slices
warnings.simplefilter("ignore", category=RuntimeWarning)
phase_stack = np.nanmean(pha_data, axis=0)
mask *= np.multiply(~np.isnan(phase_stack), phase_stack != 0.)
del phase_stack
# 1.3.3 invert pixels on mask 1+2
num_pixel2inv = int(np.sum(mask))
idx_pixel2inv = np.where(mask)[0]
print('number of pixels to invert: {} out of {} ({:.1f}%)'.format(
num_pixel2inv, num_pixel, num_pixel2inv / num_pixel * 100))
## 2. inversion
# 2.1 initiale the output matrices
ts = np.zeros((num_date, num_pixel), np.float32)
#ts_std = np.zeros((num_date, num_pixel), np.float32)
temp_coh = np.zeros(num_pixel, np.float32)
num_inv_ifg = np.zeros(num_pixel, np.int16)
# return directly if there is nothing to invert
if num_pixel2inv < 1:
ts = ts.reshape(num_date, num_row, num_col)
#ts_std = ts_std.reshape(num_date, num_row, num_col)
temp_coh = temp_coh.reshape(num_row, num_col)
num_inv_ifg = num_inv_ifg.reshape(num_row, num_col)
return ts, temp_coh, num_inv_ifg
# 2.2 un-weighted inversion (classic SBAS)
if weight_func in ['no', 'sbas']:
# a. mask for Non-Zero Phase in ALL ifgrams (share one B in sbas inversion)
if 'phase' in obs_ds_name.lower():
mask_all_net = np.all(pha_data, axis=0)
mask_all_net *= mask
else:
mask_all_net = np.array(mask)
mask_part_net = mask ^ mask_all_net
del mask
# b. invert once for all pixels with obs in all ifgrams
if np.sum(mask_all_net) > 0:
print(('inverting pixels with valid phase in all ifgrams'
' ({:.0f} pixels) ...').format(np.sum(mask_all_net)))
tsi, tcohi, num_ifgi = estimate_timeseries(
A,
B,
tbase_diff,
ifgram=pha_data[:, mask_all_net],
weight_sqrt=None,
min_norm_velocity=min_norm_velocity,
min_redundancy=min_redundancy,
skip_zero_value=skip_zero_value)
ts[:, mask_all_net] = tsi
temp_coh[mask_all_net] = tcohi
num_inv_ifg[mask_all_net] = num_ifgi
# c. pixel-by-pixel for pixels with obs not in all ifgrams
if np.sum(mask_part_net) > 0:
print(('inverting pixels with valid phase in some ifgrams'
' ({:.0f} pixels) ...').format(np.sum(mask_part_net)))
num_pixel2inv = int(np.sum(mask_part_net))
idx_pixel2inv = np.where(mask_part_net)[0]
prog_bar = ptime.progressBar(maxValue=num_pixel2inv)
for i in range(num_pixel2inv):
idx = idx_pixel2inv[i]
tsi, tcohi, num_ifgi = estimate_timeseries(
A,
B,
tbase_diff,
ifgram=pha_data[:, idx],
weight_sqrt=None,
min_norm_velocity=min_norm_velocity,
min_redundancy=min_redundancy,
skip_zero_value=skip_zero_value)
ts[:, idx] = tsi.flatten()
temp_coh[idx] = tcohi
num_inv_ifg[idx] = num_ifgi
prog_bar.update(i + 1,
every=2000,
suffix='{}/{} pixels'.format(
i + 1, num_pixel2inv))
prog_bar.close()
# 2.3 weighted inversion - pixel-by-pixel
else:
print('inverting network of interferograms into time-series ...')
prog_bar = ptime.progressBar(maxValue=num_pixel2inv)
for i in range(num_pixel2inv):
idx = idx_pixel2inv[i]
tsi, tcohi, num_ifgi = estimate_timeseries(
A,
B,
tbase_diff,
ifgram=pha_data[:, idx],
weight_sqrt=weight[:, idx],
min_norm_velocity=min_norm_velocity,
min_redundancy=min_redundancy,
skip_zero_value=skip_zero_value)
ts[:, idx] = tsi.flatten()
temp_coh[idx] = tcohi
num_inv_ifg[idx] = num_ifgi
prog_bar.update(i + 1,
every=2000,
suffix='{}/{} pixels'.format(i + 1, num_pixel2inv))
prog_bar.close()
del weight
del pha_data
## 3. prepare output
# 3.1 reshape
ts = ts.reshape(num_date, num_row, num_col)
#ts_std = ts_std.reshape(num_date, num_row, num_col)
temp_coh = temp_coh.reshape(num_row, num_col)
num_inv_ifg = num_inv_ifg.reshape(num_row, num_col)
# 3.2 convert displacement unit to meter
if obs_ds_name.startswith('unwrapPhase'):
phase2range = -1 * float(
stack_obj.metadata['WAVELENGTH']) / (4. * np.pi)
ts *= phase2range
print('converting LOS phase unit from radian to meter')
elif obs_ds_name == 'azimuthOffset':
az_pixel_size = ut.azimuth_ground_resolution(stack_obj.metadata)
ts *= az_pixel_size
print('converting azimuth offset unit from pixel ({:.2f} m) to meter'.
format(az_pixel_size))
elif obs_ds_name == 'rangeOffset':
rg_pixel_size = float(stack_obj.metadata['RANGE_PIXEL_SIZE'])
ts *= rg_pixel_size
print('converting range offset unit from pixel ({:.2f} m) to meter'.
format(rg_pixel_size))
return ts, temp_coh, num_inv_ifg, box
def simulate_coherence(date12_list, baseline_file='bl_list.txt', sensor_name='Env', inc_angle=22.8,
decor_time=200.0, coh_resid=0.2, display=False):
"""Simulate coherence for a given set of interferograms
Inputs:
date12_list - list of string in YYMMDD-YYMMDD format, indicating pairs configuration
baseline_file - string, path of baseline list text file
sensor_name - string, SAR sensor name
inc_angle - float, incidence angle
decor_time - float / 2D np.array in size of (1, pixel_num)
decorrelation rate in days, time for coherence to drop to 1/e of its initial value
coh_resid - float / 2D np.array in size of (1, pixel_num)
long-term coherence, minimum attainable coherence value
display - bool, display result as matrix or not
Output:
cohs - 2D np.array in size of (ifgram_num, pixel_num)
Example:
date12_list = pnet.get_date12_list('ifgram_list.txt')
cohs = simulate_coherences(date12_list, 'bl_list.txt', sensor_name='Tsx')
References:
Zebker, H. A., & Villasenor, J. (1992). Decorrelation in interferometric radar echoes.
IEEE-TGRS, 30(5), 950-959.
Hanssen, R. F. (2001). Radar interferometry: data interpretation and error analysis
(Vol. 2). Dordrecht, Netherlands: Kluwer Academic Pub.
Morishita, Y., & Hanssen, R. F. (2015). Temporal decorrelation in L-, C-, and X-band satellite
radar interferometry for pasture on drained peat soils. IEEE-TGRS, 53(2), 1096-1104.
Parizzi, A., Cong, X., & Eineder, M. (2009). First Results from Multifrequency Interferometry.
A comparison of different decorrelation time constants at L, C, and X Band. ESA Scientific
Publications(SP-677), 1-5.
"""
date_list, pbase_list, dop_list = read_baseline_file(baseline_file)[0:3]
tbase_list = ptime.date_list2tbase(date_list)[0]
# Thermal decorrelation (Zebker and Villasenor, 1992, Eq.4)
SNR = sensor.signal2noise_ratio(sensor_name)
coh_thermal = 1. / (1. + 1./SNR)
pbase_c = critical_perp_baseline(sensor_name, inc_angle)
bandwidth_az = sensor.azimuth_bandwidth(sensor_name)
date12_list = ptime.yyyymmdd_date12(date12_list)
ifgram_num = len(date12_list)
if isinstance(decor_time, (int, float)):
pixel_num = 1
decor_time = float(decor_time)
else:
pixel_num = decor_time.shape[1]
if decor_time == 0.:
decor_time = 0.01
cohs = np.zeros((ifgram_num, pixel_num), np.float32)
for i in range(ifgram_num):
if display:
sys.stdout.write('\rinterferogram = %4d/%4d' % (i, ifgram_num))
sys.stdout.flush()
m_date, s_date = date12_list[i].split('_')
m_idx = date_list.index(m_date)
s_idx = date_list.index(s_date)
pbase = pbase_list[s_idx] - pbase_list[m_idx]
tbase = tbase_list[s_idx] - tbase_list[m_idx]
# Geometric decorrelation (Hanssen, 2001, Eq. 4.4.12)
coh_geom = (pbase_c - abs(pbase)) / pbase_c
if coh_geom < 0.:
coh_geom = 0.
# Doppler centroid decorrelation (Hanssen, 2001, Eq. 4.4.13)
if not dop_list:
coh_dc = 1.
else:
coh_dc = calculate_doppler_overlap(dop_list[m_idx],
dop_list[s_idx],
bandwidth_az)
if coh_dc < 0.:
coh_dc = 0.
# Option 1: Temporal decorrelation - exponential delay model (Parizzi et al., 2009; Morishita and Hanssen, 2015)
coh_temp = np.multiply((coh_thermal - coh_resid), np.exp(-1*abs(tbase)/decor_time)) + coh_resid
coh = coh_geom * coh_dc * coh_temp
cohs[i, :] = coh
#epsilon = 1e-3
#cohs[cohs < epsilon] = epsilon
if display:
print('')
if display:
print(('critical perp baseline: %.f m' % pbase_c))
cohs_mat = coherence_matrix(date12_list, cohs)
plt.figure()
plt.imshow(cohs_mat, vmin=0.0, vmax=1.0, cmap='jet')
plt.xlabel('Image number')
plt.ylabel('Image number')
cbar = plt.colorbar()
cbar.set_label('Coherence')
plt.title('Coherence matrix')
plt.show()
return cohs
def simulate_coherence(date12_list,
baseline_file='bl_list.txt',
sensor_name='Env',
inc_angle=22.8,
decor_time=200.0,
coh_resid=0.2,
display=False):
"""Simulate coherence for a given set of interferograms
Inputs:
date12_list - list of string in YYMMDD-YYMMDD format, indicating pairs configuration
baseline_file - string, path of baseline list text file
sensor_name - string, SAR sensor name
inc_angle - float, incidence angle
decor_time - float / 2D np.array in size of (1, pixel_num)
decorrelation rate in days, time for coherence to drop to 1/e of its initial value
coh_resid - float / 2D np.array in size of (1, pixel_num)
long-term coherence, minimum attainable coherence value
display - bool, display result as matrix or not
Output:
cohs - 2D np.array in size of (ifgram_num, pixel_num)
Example:
date12_list = pnet.get_date12_list('ifgram_list.txt')
cohs = simulate_coherences(date12_list, 'bl_list.txt', sensor_name='Tsx')
References:
Zebker, H. A., & Villasenor, J. (1992). Decorrelation in interferometric radar echoes.
IEEE-TGRS, 30(5), 950-959.
Hanssen, R. F. (2001). Radar interferometry: data interpretation and error analysis
(Vol. 2). Dordrecht, Netherlands: Kluwer Academic Pub.
Morishita, Y., & Hanssen, R. F. (2015). Temporal decorrelation in L-, C-, and X-band satellite
radar interferometry for pasture on drained peat soils. IEEE-TGRS, 53(2), 1096-1104.
Parizzi, A., Cong, X., & Eineder, M. (2009). First Results from Multifrequency Interferometry.
A comparison of different decorrelation time constants at L, C, and X Band. ESA Scientific
Publications(SP-677), 1-5.
"""
date_list, pbase_list, dop_list = read_baseline_file(baseline_file)[0:3]
tbase_list = ptime.date_list2tbase(date_list)[0]
# Thermal decorrelation (Zebker and Villasenor, 1992, Eq.4)
SNR = sensor.signal2noise_ratio(sensor_name)
coh_thermal = 1. / (1. + 1. / SNR)
pbase_c = critical_perp_baseline(sensor_name, inc_angle)
bandwidth_az = sensor.azimuth_bandwidth(sensor_name)
date12_list = ptime.yyyymmdd_date12(date12_list)
ifgram_num = len(date12_list)
if isinstance(decor_time, (int, float)):
pixel_num = 1
decor_time = float(decor_time)
else:
pixel_num = decor_time.shape[1]
if decor_time == 0.:
decor_time = 0.01
cohs = np.zeros((ifgram_num, pixel_num), np.float32)
for i in range(ifgram_num):
if display:
sys.stdout.write('\rinterferogram = %4d/%4d' % (i, ifgram_num))
sys.stdout.flush()
m_date, s_date = date12_list[i].split('_')
m_idx = date_list.index(m_date)
s_idx = date_list.index(s_date)
pbase = pbase_list[s_idx] - pbase_list[m_idx]
tbase = tbase_list[s_idx] - tbase_list[m_idx]
# Geometric decorrelation (Hanssen, 2001, Eq. 4.4.12)
coh_geom = (pbase_c - abs(pbase)) / pbase_c
if coh_geom < 0.:
coh_geom = 0.
# Doppler centroid decorrelation (Hanssen, 2001, Eq. 4.4.13)
if not dop_list:
coh_dc = 1.
else:
coh_dc = calculate_doppler_overlap(dop_list[m_idx],
dop_list[s_idx], bandwidth_az)
if coh_dc < 0.:
coh_dc = 0.
# Option 1: Temporal decorrelation - exponential delay model (Parizzi et al., 2009; Morishita and Hanssen, 2015)
coh_temp = np.multiply(
(coh_thermal - coh_resid), np.exp(
-1 * abs(tbase) / decor_time)) + coh_resid
coh = coh_geom * coh_dc * coh_temp
cohs[i, :] = coh
#epsilon = 1e-3
#cohs[cohs < epsilon] = epsilon
if display:
print('')
if display:
print(('critical perp baseline: %.f m' % pbase_c))
cohs_mat = coherence_matrix(date12_list, cohs)
plt.figure()
plt.imshow(cohs_mat, vmin=0.0, vmax=1.0, cmap='jet')
plt.xlabel('Image number')
plt.ylabel('Image number')
cbar = plt.colorbar()
cbar.set_label('Coherence')
plt.title('Coherence matrix')
plt.show()
return cohs
