def run(self):
input_path = os.path.join(self.datapath, "input", self.input_dataset,
"images", self.filename + ".tif")
gdal_translate_args = [
'-srcwin', self.min_x, self.min_y, self.tile_width,
self.tile_height, input_path,
self.output().path
]
sh.gdal_translate(gdal_translate_args)
def gif2tif(self, start, end):
"""
Convert `prediction output`_ from GIF to TIFF.
Will use extent.json and gdal_translate to recover geospatial
information lost to GIF format.
Parameters:
- `start`: beginning of prediction range.
- `end`: ending of prediction range, inclusive.
Start and end parameters are necessary to figure out paths.
.. _`prediction output`: http://www.ncgia.ucsb.edu/projects/gig/About/dtImOut.htm
"""
predict_dir = join(self.output_path, 'predict')
with open(join(self.input_path, 'extent.json')) as extent_file:
extent = json.load(extent_file)
columnas = str(int(extent["columns"]) - 1)
renglones = str(int(extent["rows"]) - 1)
epsg = 'EPSG:' + extent["epsg"]
xmin = str(extent["xmin"])
xmax = str(extent["xmax"])
ymin = str(extent["ymin"])
ymax = str(extent["ymax"])
self.save_status('gif2tif')
for year in range(start + 1, end + 1):
gif = join(predict_dir, "%s_urban_%s.gif" % (self.location, year))
tmp_tif = join(predict_dir,
"%s_urban_%s_tmp.tif" % (self.location, year))
tmp_xml = join(
predict_dir,
"%s_urban_%s_tmp.tif.aux.xml" % (self.location, year))
tif = join(predict_dir, "%s_urban_%s.tif" % (self.location, year))
gdal_translate('-a_srs', epsg, '-ot', 'Float64', '-of', 'GTiff',
'-gcp', '0', renglones, xmin, ymin, '-gcp',
columnas, renglones, xmax, ymin, '-gcp', columnas,
'0', xmax, ymax, '-gcp', '0', '0', xmin, ymax, gif,
tmp_tif)
otbcli_BandMath('-il', tmp_tif, '-out', tif, '-exp',
'im1b1 < 9 ? im1b1 : 0')
try:
os.remove(tmp_tif)
os.remove(tmp_xml)
except:
pass
def process_layer(self, typename):
fname_in = '%s.shp' % typename
fname_out = '%s_rasterized.tiff' % typename
pwd = str(sh.pwd()).strip()
try:
sh.cd('/tmp/layer')
sh.rm("-rf",sh.glob('*'))
sh.unzip('../layer.zip')
saga_cmd.shapes_points("Points Filter", POINTS=fname_in, FIELD="MASA_HUMEDO", FILTER="tmp1.shp", RADIUS=100, MINNUM=25, MAXNUM=200, METHOD=4, PERCENT=15)
saga_cmd.shapes_points("Points Filter", POINTS="tmp1.shp", FIELD="MASA_HUMEDO", FILTER="tmp2.shp", RADIUS=100, MINNUM=25, MAXNUM=200, METHOD=5, PERCENT=90)
saga_cmd.grid_gridding("Shapes to Grid", INPUT="tmp2.shp", FIELD="MASA_HUMEDO", MULTIPLE=4, LINE_TYPE=0, GRID_TYPE=3, USER_SIZE=0.0001, TARGET=0, USER_GRID="tmp3.sgrd")
saga_cmd.grid_tools("Close Gaps", INPUT="tmp3.sgrd", RESULT="tmp4.sgrd")
saga_cmd.shapes_points("Convex Hull", SHAPES="tmp2.shp", HULLS="tmphull.shp", POLYPOINTS=0)
saga_cmd.shapes_grid("Clip Grid with Polygon", INPUT="tmp4.sgrd", OUTPUT="tmp5.sgrd", POLYGONS="tmphull.shp")
saga_cmd.grid_filter("Gaussian Filter", INPUT="tmp5.sgrd", RESULT="tmp6", SIGMA=3, MODE=1, RADIUS=50)
sh.gdal_translate("-of", "gtiff", "tmp6.sdat", fname_out)
finally:
sh.cd(pwd)
return '/tmp/layer/%s' % fname_out
def downsample(input: Path, factor: int, anti_alias: bool):
# Fail early if h5repack cli command is not available.
from sh import h5repack, gdal_translate
granule_name = find_a_granule_name(input)
fmask_image = input.with_name(f"{granule_name}.fmask.img")
nbar_size = None
with h5py.File(input) as f:
image_paths = find_h5_paths(f, "IMAGE")
for i, image_path in enumerate(image_paths):
old_image: Optional[h5py.Dataset] = f[image_path]
def info(msg: str):
secho(
f"{i: 4}/{len(image_paths)} {style(repr(image_path), fg='blue')}: {msg}"
)
old_shape = old_image.shape
if all(dim_size < factor for dim_size in old_shape):
info(f"Skipping")
continue
attrs = dict(old_image.attrs.items())
old_geotransform = attrs["geotransform"]
new_data = old_image[()][::factor, ::factor]
new_shape = new_data.shape
info(f"New shape: {new_shape!r}")
del old_image
del f[str(image_path)]
folder, name = image_path.rsplit("/", 1)
parent: h5py.Group = f[str(folder)]
image = parent.create_dataset(name, new_shape, data=new_data)
new_geotransform = list(old_geotransform)
new_geotransform[1] *= old_shape[1] / new_shape[1]
new_geotransform[5] *= old_shape[0] / new_shape[0]
attrs["geotransform"] = new_geotransform
image.attrs.update(attrs)
# Update any res group with the new resolution.
res_group_path = _get_res_group_path(image_path)
if res_group_path:
res_group = f[res_group_path]
res_group.attrs["resolution"] = [
abs(new_geotransform[5]),
abs(new_geotransform[1]),
]
if "/NBAR/" in image_path:
nbar_size = new_shape
if nbar_size is None:
raise ValueError("No nbar image found?")
# We need to repack the file to actually free up the space.
repacked = input.with_suffix(".repacked.h5")
h5repack("-f", "GZIP=5", input, repacked)
repacked.rename(input)
if fmask_image.exists():
secho(f"Scaling fmask {fmask_image}")
tmp = fmask_image.with_suffix(f".tmp.{fmask_image.suffix}")
gdal_translate("-outsize", nbar_size[1], nbar_size[0], fmask_image,
tmp)
tmp.rename(fmask_image)