def mt_layover(filelist, outfile, temp_dir, extent):
'''
This function is usally used in the time-series workflow of OST. A list
of the filepaths layover/shadow masks
:param filelist - list of files
:param out_dir - directory where the output file will be stored
:return path to the multi-temporal layover/shadow mask file generated
'''
# get the start time for Info on processing time
start = time.time()
# create path to out file
ls_layer = opj(temp_dir, os.path.basename(outfile))
# create a vrt-stack out of
print(' INFO: Creating common Layover/Shadow Mask')
vrt_options = gdal.BuildVRTOptions(srcNodata=0, separate=True)
gdal.BuildVRT(opj(temp_dir, 'ls.vrt'), filelist, options=vrt_options)
with rasterio.open(opj(temp_dir, 'ls.vrt')) as src:
# get metadata
meta = src.meta
# update driver and reduced band count
meta.update(driver='GTiff', count=1, dtype='uint8')
# create outfiles
with rasterio.open(ls_layer, 'w', **meta) as out_min:
# loop through blocks
for _, window in src.block_windows(1):
# read array with all bands
stack = src.read(range(1, src.count + 1), window=window)
# get stats
arr_max = np.nanmax(stack, axis=0)
arr = arr_max / arr_max
out_min.write(np.uint8(arr), window=window, indexes=1)
ras.mask_by_shape(ls_layer,
outfile,
extent,
to_db=False,
datatype='uint8',
rescale=False,
ndv=0)
# os.remove(ls_layer)
h.timer(start)
return outfile
def mt_extent(list_of_scenes, out_file, temp_dir, buffer=None):
out_dir = os.path.dirname(out_file)
vrt_options = gdal.BuildVRTOptions(srcNodata=0, separate=True)
# build vrt stack from all scenes
gdal.BuildVRT(opj(out_dir, 'extent.vrt'),
list_of_scenes,
options=vrt_options)
start = time.time()
outline_file = opj(temp_dir, os.path.basename(out_file))
ras.outline(opj(out_dir, 'extent.vrt'), outline_file, 0, False)
vec.exterior(outline_file, out_file, buffer)
h.delete_shapefile(outline_file)
os.remove(opj(out_dir, 'extent.vrt'))
h.timer(start)
def _grd_remove_border(infile):
'''An OST function to remove GRD border noise from Sentinel-1 data
This is a custom routine to remove GRD border noise
from Sentinel-1 GRD products. It works on the original intensity
images.
NOTE: For the common dimap format, the infile needs to be the
ENVI style file inside the *data folder.
The routine checks the outer 3000 columns for its mean value.
If the mean value is below 100, all values will be set to 0,
otherwise the routine will continue fpr another 150 columns setting
the value to 0. All further columns towards the inner image are
considered valid.
Args:
infile: string or os.path object for a
gdal compatible intensity file of Sentinel-1
Notes:
The file will be manipulated inplace, meaning,
no new outfile is created.
'''
# print(' INFO: Removing the GRD Border Noise.')
currtime = time.time()
# read raster file and get number of columns adn rows
raster = gdal.Open(infile, gdal.GA_Update)
cols = raster.RasterXSize
rows = raster.RasterYSize
# create 3000xrows array for the left part of the image
array_left = np.array(
raster.GetRasterBand(1).ReadAsArray(0, 0, 3000, rows))
for x in range(3000):
# condition if more than 50 pixels within the line have values
# less than 500, delete the line
# if np.sum(np.where((array_left[:,x] < 200)
# & (array_left[:,x] > 0) , 1, 0)) <= 50:
if np.mean(array_left[:, x]) <= 100:
array_left[:, x].fill(0)
else:
z = x + 150
if z > 3000:
z = 3000
for y in range(x, z, 1):
array_left[:, y].fill(0)
cols_left = y
break
try:
cols_left
except NameError:
cols_left = 3000
# write array_left to disk
# print(' INFO: Total amount of columns: {}'.format(cols_left))
# print(' INFO: Number of colums set to 0 on the left side: '
# ' {}'.format(cols_left))
# raster.GetRasterBand(1).WriteArray(array_left[:, :+cols_left], 0, 0, 1)
raster.GetRasterBand(1).WriteArray(array_left[:, :+cols_left], 0, 0)
array_left = None
# create 2d array for the right part of the image (3000 columns and rows)
cols_last = cols - 3000
array_right = np.array(
raster.GetRasterBand(1).ReadAsArray(cols_last, 0, 3000, rows))
# loop through the array_right columns in opposite direction
for x in range(2999, 0, -1):
if np.mean(array_right[:, x]) <= 100:
array_right[:, x].fill(0)
else:
z = x - 150
if z < 0:
z = 0
for y in range(x, z, -1):
array_right[:, y].fill(0)
cols_right = y
break
try:
cols_right
except NameError:
cols_right = 0
#
col_right_start = cols - 3000 + cols_right
# print(' INFO: Number of columns set to 0 on the'
# ' right side: {}'.format(3000 - cols_right))
# print(' INFO: Amount of columns kept: {}'.format(col_right_start))
raster.GetRasterBand(1).WriteArray(array_right[:, cols_right:],
col_right_start, 0)
array_right = None
h.timer(currtime)
def grd_remove_border(infile):
"""OST function to remove GRD border noise from Sentinel-1 data
This is a custom routine to remove GRD border noise
from Sentinel-1 GRD products. It works on the original intensity
images.
NOTE: For the common dimap format, the infile needs to be the
ENVI style file inside the *data folder.
The routine checks the outer 3000 columns for its mean value.
If the mean value is below 100, all values will be set to 0,
otherwise the routine will continue fpr another 150 columns setting
the value to 0. All further columns towards the inner image are
considered valid.
:param infile:
:return:
"""
logger.debug(f'Removing the GRD Border Noise for {infile.name}.')
currtime = time.time()
# read raster file and get number of columns adn rows
raster = gdal.Open(str(infile), gdal.GA_Update)
cols = raster.RasterXSize
rows = raster.RasterYSize
# create 3000xrows array for the left part of the image
array_left = np.array(
raster.GetRasterBand(1).ReadAsArray(0, 0, 3000, rows))
cols_left = 3000
for x in range(3000):
# if mean value of column is below 100, fil with 0s
if np.mean(array_left[:, x]) <= 100:
array_left[:, x].fill(0)
else:
z = x + 200
if z > 3000:
z = 3000
for cols_left in range(x, z, 1):
array_left[:, cols_left].fill(0)
break
# write array_left to disk
logger.debug(f'Number of colums set to 0 on the left side: {cols_left}.')
raster.GetRasterBand(1).WriteArray(array_left[:, :+cols_left], 0, 0)
del array_left
# create 2d array for the right part of the image (3000 columns and rows)
cols_last = cols - 3000
array_right = np.array(
raster.GetRasterBand(1).ReadAsArray(cols_last, 0, 3000, rows))
# loop through the array_right columns in opposite direction
cols_right = 3000
for x in range(2999, 0, -1):
if np.mean(array_right[:, x]) <= 100:
array_right[:, x].fill(0)
else:
z = x - 200
if z < 0:
z = 0
for cols_right in range(x, z, -1):
array_right[:, cols_right].fill(0)
break
col_right_start = cols - 3000 + cols_right
logger.debug(
f'Number of columns set to 0 on the right side: {3000 - cols_right}.')
logger.debug(f'Amount of columns kept: {col_right_start}.')
raster.GetRasterBand(1).WriteArray(array_right[:, cols_right:],
col_right_start, 0)
logger.debug(h.timer(currtime))
def _timeseries_to_timescan(burst_inventory, processing_dir, temp_dir,
burst_dir, to_db, metrics, outlier_removal):
product_list = [
'BS.HH', 'BS.VV', 'BS.HV', 'BS.VH', 'coh.VV', 'coh.VH', 'Alpha',
'Entropy', 'Anisotropy'
]
for product in product_list:
for timeseries in glob.glob(
opj(burst_dir, 'Timeseries', '*{}*vrt'.format(product))):
print(' INFO: Creating timescan for {}'.format(product))
timescan_dir = opj(burst_dir, 'Timescan')
os.makedirs(timescan_dir, exist_ok=True)
# we get the name of the time-series parameter
polarisation = timeseries.split('/')[-1].split('.')[2]
if polarisation == 'vrt':
timescan_prefix = opj(
'{}'.format(timescan_dir),
'{}'.format(timeseries.split('/')[-1].split('.')[1]))
else:
timescan_prefix = opj(
'{}'.format(timescan_dir), '{}.{}'.format(
timeseries.split('/')[-1].split('.')[1], polarisation))
start = time.time()
if 'BS.' in timescan_prefix: # backscatter
ts.mt_metrics(timeseries,
timescan_prefix,
metrics,
rescale_to_datatype=True,
to_power=to_db,
outlier_removal=outlier_removal)
else: # non-backscatter
ts.mt_metrics(timeseries,
timescan_prefix,
metrics,
rescale_to_datatype=False,
to_power=False,
outlier_removal=outlier_removal)
h.timer(start)
# rename and create vrt
# print('renaming')
i, list_of_files = 0, []
for product in itertools.product(product_list, metrics):
file = glob.glob(
opj(burst_dir, 'Timescan',
'*{}.{}.tif'.format(product[0], product[1])))
if file:
i += 1
outfile = opj(burst_dir, 'Timescan',
'{}.{}.{}.tif'.format(i, product[0], product[1]))
shutil.move(file[0], outfile)
list_of_files.append(outfile)
# create vrt
vrt_options = gdal.BuildVRTOptions(srcNodata=0, separate=True)
gdal.BuildVRT(opj(burst_dir, 'Timescan', 'Timescan.vrt'),
list_of_files,
options=vrt_options)
def mt_layover(filelist, outfile, temp_dir, extent, update_extent=False):
'''
This function is usally used in the time-series workflow of OST. A list
of the filepaths layover/shadow masks
:param filelist - list of files
:param out_dir - directory where the output file will be stored
:return path to the multi-temporal layover/shadow mask file generated
'''
if type(filelist) == str:
filelist = filelist.replace("'", '').strip('][').split(', ')
if type(update_extent) == str:
if update_extent == 'False':
update_extent = False
# get some info
burst_dir = os.path.dirname(outfile)
burst = os.path.basename(burst_dir)
extent = opj(burst_dir, '{}.extent.shp'.format(burst))
# get the start time for Info on processing time
start = time.time()
# create path to out file
ls_layer = opj(temp_dir, os.path.basename(outfile))
# create a vrt-stack out of
print(' INFO: Creating common Layover/Shadow Mask')
vrt_options = gdal.BuildVRTOptions(srcNodata=0, separate=True)
gdal.BuildVRT(opj(temp_dir, 'ls.vrt'), filelist, options=vrt_options)
with rasterio.open(opj(temp_dir, 'ls.vrt')) as src:
# get metadata
meta = src.meta
# update driver and reduced band count
meta.update(driver='GTiff', count=1, dtype='uint8')
# create outfiles
with rasterio.open(ls_layer, 'w', **meta) as out_min:
# loop through blocks
for _, window in src.block_windows(1):
# read array with all bands
stack = src.read(range(1, src.count + 1), window=window)
# get stats
arr_max = np.nanmax(stack, axis=0)
arr = arr_max / arr_max
out_min.write(np.uint8(arr), window=window, indexes=1)
ras.mask_by_shape(ls_layer,
outfile,
extent,
to_db=False,
datatype='uint8',
rescale=False,
ndv=0)
os.remove(ls_layer)
h.timer(start)
if update_extent:
print(' INFO: Calculating symetrical difference of extent and ls_mask')
# polygonize the multi-temporal ls mask
ras.polygonize_raster(outfile, '{}.shp'.format(outfile[:-4]))
# create file for masked extent
extent_ls_masked = opj(burst_dir, '{}.extent.masked.shp'.format(burst))
# calculate difference between burst exntetn and ls mask, fr masked extent
vec.difference(extent, '{}.shp'.format(outfile[:-4]), extent_ls_masked)