def getDataset(scene_path, pol):
ds = None
transform = {}
# generate file list
filelist = fio.getFileList(
'S1*_*{}_tnr_bnr_Orb_Cal_ML_TF_TC_dB.tif'.format(pol), scene_path)
if len(filelist) == 1:
# open raster
ds = gdal.Open(filelist[0])
if ds is not None:
transform['fwd'] = affine.Affine.from_gdal(*ds.GetGeoTransform())
transform['rev'] = ~transform['fwd']
band = ds.GetRasterBand(1)
return {'ds': ds, 'band': band, 'transform': transform}
def compileVrt(scene, product):
out_pathname = None
# generate file list
filelist = fio.getFileList('*', scene)
if len(filelist) > 0:
# need to guarantee consistent band ordering
sorted_list = []
# for each band
bands = product.getElementsByTagName('band')
for band in bands:
if (band.hasAttribute("filename")):
for obj in filelist:
# add entry to sort list if configuration matches argument
if fnmatch.fnmatch(os.path.basename(obj),
str(band.attributes["filename"].value)):
sorted_list.append(obj)
break
# build vrt on validation of successful sort
if len(sorted_list) == len(bands):
out_pathname = scene + '/' + product.attributes["name"].value + '.vrt'
updateImages(sorted_list)
vrt = gdal.BuildVRT(out_pathname,
sorted_list,
options=gdal.BuildVRTOptions(separate=True))
vrt = None
return out_pathname
def checkScene(pathname):
sceneOK = False
# decompress product file
path = os.path.dirname(pathname)
out, err, code = ps.execute('unzip', ["-o", "-d", path, pathname])
if (code <= 1):
filelist = fio.getFileList('*.tiff', path)
if len(filelist) > 0:
# open scene and extract gcps
in_ds = gdal.Open(filelist[0])
gcps = in_ds.GetGCPs()
min_x = 180
max_x = -180
for gcp in gcps:
min_x = min(min_x, gcp.GCPX)
max_x = max(max_x, gcp.GCPX)
# large longitude difference when crossing antimeridian
if max_x - min_x < 10:
sceneOK = True
else:
print('... scene crosses antimeridan - skipping: {}'.format(
pathname))
# housekeeping of raw zip sub-folder
zip_path = os.path.splitext(pathname)[0]
if (os.path.exists(zip_path)):
shutil.rmtree(zip_path)
return sceneOK
def getSceneList(args):
scene_list = []
# single file argument
if args.file is not None:
scene_list.append(args.file)
else:
# generate file list
filelist = fio.getFileList('*.zip', args.path)
if len(filelist) > 0:
# parse datetime from pathname
for obj in filelist:
dt = parser.getDateTime(obj)
# get files satisfying constraints
if dt.timestamp() >= args.start.timestamp() and dt.timestamp(
) <= args.end.timestamp():
scene_list.append(obj)
return sorted(scene_list)
out_path = os.path.join(ard_path, args.product)
# geocode scene
geocode(
scene=scene,
dem=args.dem,
tempdir=os.path.join(out_path, 'process'),
outdir=out_path,
targetres=args.res,
scaling='db',
func_geoback=1,
# cleanup=False,
export_extra=['inc_geo', 'ls_map_geo'])
# reproject imagery to epsg:3460
image_list = fio.getFileList('S1*.tif', out_path)
options = reproject.getTransform(image_list[0], {
't_epsg': 32760,
'res_x': 20,
'res_y': 20
})
for img_pathname in image_list:
warp_pathname = img_pathname.replace('.tif', '_warp.tif')
reproject.toEpsg(
img_pathname, warp_pathname, options,
['TILED=YES', 'BLOCKXSIZE=256', 'BLOCKYSIZE=256'])
else:
print('scene crosses anti-meridian: ' + scene)
plt.ylabel("relative information")
plt.xlabel("features")
plt.show()
# compute svm fit
#model = SVC( kernel='rbf' )
model = MLPClassifier(hidden_layer_sizes=(30, 30, 30), max_iter=1000)
model.fit(X, y)
y_pred = model.predict(X)
print(confusion_matrix(y, y_pred))
print(classification_report(y, y_pred))
# generate file list
filelist = fio.getFileList('S1*_VV_tnr_bnr_Orb_Cal_ML_TF_TC_dB.tif',
'/data/ard/fiji')
for obj in filelist:
path = os.path.dirname(obj)
# get training set
feature, name_list = getFeatureFrame(sample, path, 1000)
# get scatter matrix
X, y = getScatterMatrix(feature)
#plotFeatureSeparability( X, y, name_list )
y_pred = model.predict(X)
print(path)
print(confusion_matrix(y, y_pred))