def process(self,
scene_path,
aois,
out_path=None,
distance=100,
ext='*_reflectance.tif'):
"""
Placeholder
"""
aoi_paths = []
# get exported geotiff list
files = getFileList(scene_path, ext)
for f in files:
# open image
ds = gdal.Open(f)
if ds is not None:
# get aoi to image transform
extent = self.getExtent(ds)
coord_tx = self.getCoordinateTransform(ds)
# for each aoi
for aoi in aois:
# get buffered bounding box coordinates
bbox = self.getBoundingBox(aoi['bbox'],
coord_tx['aoi_image'],
distance=distance)
if self.overlapsScene(extent, bbox) is True:
# create aoi sub-path
aoi_path = os.path.join(scene_path, aoi['name'] + '/')
if not os.path.exists(aoi_path):
os.makedirs(aoi_path, 0o755)
# generate aoi sub-image aligned with bbox
aoi_pathname = os.path.join(aoi_path,
os.path.basename(f))
print('Creating AoI image: {}'.format(aoi_pathname))
# reproject bbox to local utm - setup warp options
bbox = self.getBoundingBox(aoi['bbox'],
coord_tx['aoi_local'],
distance=distance)
options = '-t_srs EPSG:{} -tr 15 -15 -te {} {} {} {}'.format(
self._epsg, bbox['ulx'], bbox['lry'], bbox['lrx'],
bbox['uly'])
gdal.Warp(aoi_pathname, ds, options=options)
# record aoi image location
aoi_paths.append(aoi_path)
return list(set(aoi_paths))
def process(self, scenes, bounds, out_path):
"""
placeholder
"""
# get client connection to s2 bucket
client = storage.Client()
bucket = client.get_bucket("gcp-public-data-sentinel-2")
for idx, row in scenes.iterrows():
# next interation if no intersection
if row['east_lon'] < bounds[0] or row['west_lon'] > bounds[
2] or row['south_lat'] > bounds[1] or row[
'north_lat'] < bounds[3]:
continue
# evaluate number of successfully downloaded scenes
files = getFileList(out_path, 'manifest.safe')
if len(files) > self._max_scenes:
# maximum chips found
print('maximum scenes {} downloaded: {}'.format(
self._max_scenes, out_path))
break
# construct scene path
dt = datetime.strptime(row['sensing_time'],
'%Y-%m-%dT%H:%M:%S.%fZ')
scene_path = os.path.join(
os.path.join(out_path, dt.strftime('%Y%m%d_%H%M%S')),
row['mgrs_tile'])
# get filelist in bucket prefix subfolder
prefix = row['base_url'][len('gs://gcp-public-data-sentinel-2/'):]
blobs = bucket.list_blobs(prefix=prefix, delimiter=None)
for blob in blobs:
# apply match to bucket files
d = self.blob2dict(blob)
if any(f in d['name'] for f in self._prefixes):
if not os.path.exists(scene_path):
os.makedirs(scene_path)
print('downloading {} -> {}'.format(d['name'], scene_path))
# download bucket file to local directory
pathname = os.path.join(scene_path,
os.path.basename(d['name']))
if not os.path.exists(pathname):
with open(pathname, 'w+b') as z:
blob.download_to_file(z)
return
def getSceneList(args):
"""
Placeholder
"""
# assume single scene - else collect list
scenes = [args.scene]
if args.batch is True:
scenes = getFileList(args.scene, 'S2*.zip')
return scenes
def getMosaicDataset(self, path, search):
"""
Placeholder
"""
# return package
dataset = {'srs': None, 'channels': []}
# load band data into list
files = sorted(getFileList(path, search))
for f in files:
# get band index
match = re.search('B[0-9]{2}', f)
if match:
# load geotiff
ds = gdal.Open(f)
if ds is not None:
# get srs attributes
if dataset['srs'] is None:
dataset['srs'] = {
'geo': ds.GetGeoTransform(),
'prj': ds.GetProjection()
}
# create dictionary entry
dataset['channels'].append({
'index':
int(''.join(filter(str.isdigit, match.group(0)))),
'data':
ds.GetRasterBand(1).ReadAsArray(),
})
return dataset
def process(self, scene_path, aois, out_path=None, distance=100):
"""
Placeholder
"""
aoi_paths = []
# get exported geotiff list
files = getFileList(scene_path, '*_B*_*m.tif')
for f in files:
# open image
ds = gdal.Open(f)
if ds is not None:
# get aoi to image transform
extent = self.getExtent(ds)
coord_tx = self.getCoordinateTransform(ds)
# for each aoi
for aoi in aois:
# get buffered bounding box coordinates
bbox = self.getBoundingBox(aoi['bbox'],
coord_tx['aoi_image'],
distance=distance)
"""
ulx, uly, ulz = coord_rx.TransformPoint( bbox[ 'lrx' ], bbox[ 'lry' ] )
lrx, lry, lrz = coord_rx.TransformPoint( bbox[ 'ulx' ], bbox[ 'uly' ] )
print ( aoi[ 'name' ], ulx, uly, lrx, lry )
"""
if self.overlapsScene(extent, bbox) is True:
# create aoi sub-path
aoi_path = os.path.join(scene_path, aoi['name'] + '/')
if not os.path.exists(aoi_path):
os.makedirs(aoi_path, 0o755)
# generate aoi sub-image aligned with bbox
aoi_pathname = os.path.join(aoi_path,
os.path.basename(f))
print('Creating AoI image: {}'.format(aoi_pathname))
# reproject bbox to local utm and fix pixel resolution
bbox = self.getBoundingBox(aoi['bbox'],
coord_tx['aoi_local'],
distance=distance)
res_option = self.getResolution(os.path.basename(f))
# reproject bbox to local utm - setup warp options
options = '-t_srs EPSG:{} -tr {} -te {} {} {} {}'.format(
self._epsg, res_option, bbox['ulx'], bbox['lry'],
bbox['lrx'], bbox['uly'])
gdal.Warp(aoi_pathname, ds, options=options)
# resample 20m resolution sub-image to 10m
if '20m' in aoi_pathname:
print('Creating resampled AoI image: {}'.format(
aoi_pathname.replace('20m', '10m')))
# rerun gdalwarp
options = '-t_srs EPSG:{} -tr 10 -10 -te {} {} {} {}'.format(
self._epsg, bbox['ulx'], bbox['lry'],
bbox['lrx'], bbox['uly'])
gdal.Warp(aoi_pathname.replace('20m', '10m'),
ds,
options=options)
# record aoi image location
aoi_paths.append(aoi_path)
return list(set(aoi_paths))
def getImageChip(self,
path,
centroid,
out_pathname,
size=512,
overwrite=False):
"""
Placeholder
"""
# file not created or overwrite
if not os.path.exists(out_pathname):
# get scenes downloaded for datetime
images = self.clipScenes(getFileList(path, '_TCI.jp2'), centroid,
size)
# images = [ '/data/test/ALB0003/20170820_093607/34TDL/T34TDL_20170820T093041_TCI-chip.tif' ]
if len(images) > 0:
# get scene srs
ds = gdal.Open(images[0])
coord_tx = self.getCoordinateTransform(ds)
# transform latlon centroid coordinates to image srs
x, y, z = coord_tx.TransformPoint(centroid[0], centroid[1])
x = round(x)
y = round(y)
distance = (size / 2) * self._resolution
x0 = x - distance
y0 = y - distance
x1 = x + distance
y1 = y + distance
print(x0, y0, x1, y1)
# compile translation options
options = [
'-of MEM', '-t_srs {}'.format(ds.GetProjection()),
'-te {} {} {} {}'.format(x0, y0, x1, y1),
'-tr {} {}'.format(self._resolution,
self._resolution), '-srcnodata 0'
]
# combine images into final output
options_str = ' '.join(options)
warp_ds = gdal.Warp('', images, options=options_str)
# validate goodness - number of black (masked) pixels
goodness = (
np.count_nonzero(warp_ds.GetRasterBand(1).ReadAsArray()) /
(size * size)) * 100.0
if goodness > 95.0:
# spit in-memory image to jpg file
if not os.path.exists(os.path.dirname(out_pathname)):
os.makedirs(os.path.dirname(out_pathname))
gdal.Translate(out_pathname, warp_ds, options='-of JPEG')
print('Generated chip: {} {}'.format(
out_pathname, goodness))
else:
# failed qc check
print('Chip failed QC: {} {}'.format(
out_pathname, goodness))
# housekeeping
warp_ds = None
ds = None
return
