def fmask(self):
'''-----\n
This method applies Fmask algorithm or its equivalent with the BQA band (default for Ecopotential)'''
os.chdir(self.ruta_escena)
print('Starting Cloud Mask')
try:
print('Starting Fmask')
t = time.time()
#Last value is the cloud confidence value
a = os.system(
'removethisisFmaskinstalled/usr/GERS/Fmask_4_0/application/run_Fmask_4_0.sh /usr/local/MATLAB/MATLAB_Runtime/v93 3 3 1 {}'
.format(self.umbral))
a
if a == 0:
self.cloud_mask = 'Fmask'
print('Cloud Mask (Fmask) generated in ' +
str(t - time.time()) + ' seconds')
else:
print('Starting Cloud Mask with BQA band')
outfile = os.path.join(
os.path.join(self.ruta_escena, self.escena + '_Fmask.tif'))
for i in os.listdir(self.ruta_escena):
if i.endswith('BQA.TIF'):
bqa = os.path.join(self.ruta_escena, i)
masker = LandsatMasker(bqa, collection=1)
conf = LandsatConfidence.high
cloud = masker.get_cloud_mask(conf)
cirrus = masker.get_cirrus_mask(conf)
shadow = masker.get_cloud_shadow_mask(conf)
MASCARA = cloud + cirrus + shadow
masker.save_tif(MASCARA, outfile)
self.cloud_mask = 'BQA'
print('Cloud Mask (BQA) generated in ' + str(t - time.time()) +
' seconds')
print('Calling get_water')
self.get_water()
except Exception as e:
print("Unexpected error:", type(e), e)
def img_preprocess(qa, b4, b3, b2, tdimn):
masker = LandsatMasker(qa, collection=1)
conf = LandsatConfidence.high
mask1 = masker.get_cirrus_mask(conf)
mask2 = masker.get_cloud_mask(conf)
mask = np.bitwise_or(mask1, mask2)
mask = misc.imresize(mask, tdimn)
#mask = misc.imrotate(mask,18,interp='nearest')
img1 = gdal.Open(b4)
orig1 = img1.ReadAsArray()
orig1 = misc.imresize(orig1, tdimn)
orig1 = orig1[:, :, np.newaxis]
img2 = gdal.Open(b3)
orig2 = img2.ReadAsArray()
orig2 = misc.imresize(orig2, tdimn)
orig2 = orig2[:, :, np.newaxis]
img3 = gdal.Open(b2)
orig3 = img3.ReadAsArray()
orig3 = misc.imresize(orig3, tdimn)
orig3 = orig3[:, :, np.newaxis]
orig = np.concatenate((orig1, orig2, orig3), axis=2)
#orig = misc.imrotate(orig,18,interp='nearest')
#io.imsave(os.path.join(path,'rgb2.png'),orig)
#img = io.imread(os.path.join(path,'rgb2.png'))
# =============================================================================
# plt.figure(figsize=(40,20))
# plt.subplot(131)
# plt.imshow(orig)
# plt.title('Cloud-1')
#
# plt.figure(figsize=(40,20))
# plt.subplot(131)
# plt.imshow(np.squeeze(orig1,axis=2),cmap = plt.get_cmap('gist_gray'))
# plt.imshow(mask, cmap = plt.get_cmap('Reds'), alpha=0.5)
# plt.title('Cloud-2')
# =============================================================================
orig = np.reshape(orig, (1, mask.shape[0], mask.shape[1], 3))
img_m = np.reshape(mask, (1, mask.shape[0], mask.shape[1], 1))
return (orig, img_m)
def img_preprocess(qa,b4,b3,b2, pt):
masker = LandsatMasker(qa,collection=1)
conf = LandsatConfidence.high
mask1 = masker.get_cirrus_mask(conf)
mask2 = masker.get_cloud_mask(conf)
mask = np.bitwise_or(mask1,mask2)
mask = misc.imresize(mask,tdimn)
img1 = gdal.Open(b4)
orig1 = img1.ReadAsArray()
orig1 = misc.imresize(orig1,tdimn)
orig1 = orig1[:,:,np.newaxis]
img2 = gdal.Open(b3)
orig2 = img2.ReadAsArray()
orig2 = misc.imresize(orig2,tdimn)
orig2 = orig2[:,:,np.newaxis]
img3 = gdal.Open(b2)
orig3 = img3.ReadAsArray()
orig3 = misc.imresize(orig3,tdimn)
orig3 = orig3[:,:,np.newaxis]
orig = np.concatenate((orig1,orig2,orig3),axis=2)
# =============================================================================
# io.imsave(os.path.join(pt,'rgb2.png'),orig)
# img_t = load_img(os.path.join(pt,'rgb2.png'),grayscale=True)
# img_t = img_to_array(img_t)
# =============================================================================
# =============================================================================
# plt.figure(figsize=(40,20))
# plt.subplot(131)
# plt.imshow(orig,cmap = plt.get_cmap('gist_gray'))
# plt.imshow(mask, cmap = plt.get_cmap('Reds'), alpha=0.3)
# plt.title('Cloud-2')
# plt.show()
# =============================================================================
img_m = np.reshape(mask,(mask.shape[0],mask.shape[1],1))
#print(img_t.shape,img_m.shape)
return (orig, img_m)
def img_preprocess(i, qa):
masker = LandsatMasker(qa, collection=1)
conf = LandsatConfidence.high
mask1 = masker.get_cirrus_mask(conf)
mask2 = masker.get_cloud_mask(conf)
mask = np.bitwise_or(mask1, mask2)
mask = misc.imresize(mask, tdimn)
orig = misc.imread(os.path.join(dir1, "All_Imgs/orig_{}.png".format(i)))
# =============================================================================
# plt.figure(figsize=(40,20))
# plt.subplot(131)
# plt.imshow(orig)
# plt.imshow(mask, cmap = plt.get_cmap('Reds'), alpha=0.3)
# plt.title('Cloud-2')
# plt.show()
# =============================================================================
img_m = np.reshape(mask, (mask.shape[0], mask.shape[1], 1))
return (orig, img_m)
def get_cloud_mask(bqa):
'''
Takes in a bqa array and returns a mask of clouds
'''
landsat_confidences = {
LandsatConfidence.low: 1,
LandsatConfidence.medium: 2,
LandsatConfidence.high: 3
}
bqa = bqa.astype(np.int16)
conf_is_cumulative = False
masker = LandsatMasker(bqa, collection=0)
cloud_conf = np.zeros_like(bqa).astype(np.int16)
for conf in landsat_confidences:
mask = masker.get_cloud_mask(conf, cumulative=conf_is_cumulative)
cloud_conf += mask * landsat_confidences[
conf] # multiply each conf by a value
return cloud_conf
import tensorflow as tf
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense
from keras import optimizers
from keras.utils import to_categorical
import os
#%%
path = os.path.dirname(__file__)
tdimn = (256, 256)
masker = LandsatMasker('LC08_L1TP_210017_20170711_20170711_01_RT_BQA.TIF',
collection=1)
conf = LandsatConfidence.high
mask1 = masker.get_cirrus_mask(conf)
mask2 = masker.get_cloud_mask(conf)
mask = np.bitwise_or(mask1, mask2)
mask = misc.imresize(mask, tdimn)
mask = misc.imrotate(mask, 18, interp='nearest')
#%%
img1 = gdal.Open("LC08_L1TP_210017_20170711_20170711_01_RT_B4.TIF")
orig1 = img1.ReadAsArray()
orig1 = misc.imresize(orig1, tdimn)
orig1 = orig1[:, :, np.newaxis]
img2 = gdal.Open("LC08_L1TP_210017_20170711_20170711_01_RT_B3.TIF")
orig2 = img2.ReadAsArray()
orig2 = misc.imresize(orig2, tdimn)
orig2 = orig2[:, :, np.newaxis]
output_shadow_mask_path = "{0}/{1}/{1}_bqa_shadow_mask.TIF".format(
landsat_scenes_path, sceneID)
output_cloudshadow_mask_path = "{0}/{1}/{1}_bqa_cloudshadow_mask.TIF".format(
landsat_scenes_path, sceneID)
# load the QA band directly
#
# The "collection" parameter is required for landsat to specify the collection
# number. Acceptable number: 0 (pre-collection), 1 (collection-1)
masker = LandsatMasker(bqa_band_path, collection=1)
# algorithm has high confidence that this condition exists
# (67-100 percent confidence)
conf = LandsatConfidence.medium
# Get mask indicating cloud pixels with high confidence
cloud_mask = masker.get_cloud_mask(conf, cumulative=True)
# save the result
masker.save_tif(cloud_mask, output_cloud_mask_path)
if shadow:
# Get mask indicating cloud pixels with high confidence
cloud_shadow_mask = masker.get_cloud_shadow_mask(conf, cumulative=True)
masker.save_tif(cloud_shadow_mask, output_shadow_mask_path)
# Merge the Cloud and Cloud-Shadow mask
gdal_calc_string = "gdal_calc.py -A {} -B {} --outfile={} --calc='logical_or(A==1,B==1)'".format(
output_cloud_mask_path, output_shadow_mask_path,
output_cloudshadow_mask_path)
os.system(gdal_calc_string)
# Done
