def _getDatatype(driver):
tnames = tuple(
driver.GetMetadata_Dict()["DMD_CREATIONDATATYPES"].split(" "))
types = tuple(gdal.GetDataTypeByName(t) for t in tnames)
tdict = tuple((gdal.GetDataTypeSize(t), t) for t in types)
otype = max(tdict, key=lambda x: x[0])[-1]
return np.dtype(_TYPEMAP[otype])
def create_paux_header(self,filename,datatype):
(path, ext) = os.path.splitext(filename)
auxname = path + ".aux"
if os.path.isfile(auxname):
resp = GtkExtra.message_box('Confirmation', \
auxname + ' exists. Overwrite?', ('Yes','No'))
if resp == 'No':
return
# Take the image parameters
header = long(self.header_entry.get_text())
width = long(self.width_entry.get_text())
height = long(self.height_entry.get_text())
bands = long(self.bands_entry.get_text())
aux_type_dict={'Byte':'8U','Int16':'16S','UInt16':'16U',
'Float32':'32R'}
aux_type_list = ['8U', '16S', '16U', '32R']
type = aux_type_dict[datatype]
gdaltype = gdal.GetDataTypeByName(datatype)
interleaving = self.interleave_list[self.interleave_menu.get_history()]
datasize = gdal.GetDataTypeSize(gdaltype) / 8
# Calculate offsets
pixel_offset = []
line_offset = []
image_offset = []
if interleaving is 'Pixel':
for i in range(bands):
pixel_offset.append(datasize * bands)
line_offset.append(datasize * width * bands)
image_offset.append(header + datasize * i)
elif interleaving is 'Band':
for i in range(bands):
pixel_offset.append(datasize)
line_offset.append(datasize * width)
image_offset.append(header + datasize * width * height * i)
elif interleaving is 'Line':
for i in range(bands):
pixel_offset.append(datasize)
line_offset.append(datasize * width * bands)
image_offset.append(header + datasize * width * i)
else:
raise 'Unsupported interleaving type!'
aux_swap_list = ['Swapped', 'Unswapped']
swap = aux_swap_list[self.swap_menu.get_history()]
# Write out the auxilary file
aux = open(auxname, "wt")
aux.write("AuxilaryTarget: " + os.path.basename(filename) + '\n')
aux.write("RawDefinition: " + str(width) + ' ' \
+ str(height) + ' ' + str(bands) + '\n')
for i in range(bands):
aux.write("ChanDefinition-" + str(i + 1) + ': ' + type + ' ' \
+ str(image_offset[i]) + ' ' + str(pixel_offset[i]) \
+ ' ' + str(line_offset[i]) + ' ' + swap + '\n')
aux.close()
aux = None
def _load_image(self):
# Loads the optical and segmented image data from disk. Should only be called from
# load_next_image method.
full_image_name = self.available_images[self.im_index]
self.im_name = os.path.splitext(os.path.split(full_image_name)[1])[0]
src_ds = gdal.Open(full_image_name, gdal.GA_ReadOnly)
# Read the image date from the metadata
metadata = src_ds.GetMetadata()
self.im_date = pp.parse_metadata(metadata, self.im_type)
# Determine the datatype
src_dtype = gdal.GetDataTypeSize(src_ds.GetRasterBand(1).DataType)
# Calculate the reference points from the image histogram
lower, upper, wb_ref, br_ref = pp.histogram_threshold(
src_ds, src_dtype)
self.wb_ref = np.array(wb_ref, dtype=c_uint8)
self.br_ref = np.array(br_ref, dtype=c_uint8)
# Load the image data
image_data = src_ds.ReadAsArray()
# Close the GDAL dataset
src_ds = None
# Rescale the input dataset using a histogram stretch
image_data = pp.rescale_band(image_data, lower, upper)
# Apply a white balance to the image
image_data = pp.white_balance(image_data, self.wb_ref.astype(np.float),
float(np.amax(self.wb_ref)))
# Convert the input data to c_uint8
self.original_image = np.ndarray.astype(image_data, c_uint8)
print("Creating segments on provided image...")
watershed_image = segment_image(image_data, image_type=self.im_type)
# Convert the segmented image to c_int datatype. This is needed for the
# Cython methods that calculate attribute of segments.
self.segmented_image = np.ndarray.astype(watershed_image, c_uint32)
# Clear these from memory explicitly
image_data = None
watershed_image = None
def _calculate_scale_offset(nodata, band):
"""
This method comes from the old ULA codebase.
"""
nbits = gdal.GetDataTypeSize(band.DataType)
dfScaleDstMin, dfScaleDstMax = 0.0, 255.0
if nbits == 16:
count = 32767 + nodata
histogram = band.GetHistogram(-32767, 32767, 65536)
else:
count = 0
histogram = band.GetHistogram()
dfScaleSrcMin = count
total = 0
cliplower = int(0.01 * (sum(histogram) - histogram[count]))
clipupper = int(0.99 * (sum(histogram) - histogram[count]))
while total < cliplower and count < len(histogram) - 1:
count += 1
total += int(histogram[count])
dfScaleSrcMin = count
if nbits == 16:
count = 32767 + nodata
else:
count = 0
total = 0
dfScaleSrcMax = count
while total < clipupper and count < len(histogram) - 1:
count += 1
total += int(histogram[count])
dfScaleSrcMax = count
if nbits == 16:
dfScaleSrcMin -= 32768
dfScaleSrcMax -= 32768
# Determine gain and offset
diff_ = dfScaleSrcMax - dfScaleSrcMin
# From the old Jobmanager codebase: avoid divide by zero caused by some stats.
if diff_ == 0:
_LOG.warning("dfScaleSrc Min and Max are equal! Applying correction")
diff_ = 1
dfScale = (dfScaleDstMax - dfScaleDstMin) / diff_
dfOffset = -1 * dfScaleSrcMin * dfScale + dfScaleDstMin
return dfScale, dfOffset
def create_vrt_lines(self,filename):
image_offset = long(self.header_entry.get_text())
width = long(self.width_entry.get_text())
height = long(self.height_entry.get_text())
bands = long(self.bands_entry.get_text())
interleaving = self.interleave_list[self.interleave_menu.get_history()]
dtype = self.type_list[self.type_menu.get_history()]
gdaltype = gdal.GetDataTypeByName(dtype)
datasize = gdal.GetDataTypeSize(gdaltype) / 8
byteorder = ['LSB','MSB'][self.swap_menu.get_history()]
vrtdsc = vrtutils.VRTDatasetConstructor(width,height)
if interleaving == 'Pixel':
pixoff = datasize*bands
lineoff = pixoff*width
for idx in range(bands):
imoff = image_offset + idx*datasize
vrtdsc.AddRawBand(filename, dtype, byteorder,
imoff, pixoff, lineoff)
elif interleaving == 'Line':
pixoff=datasize
lineoff=datasize*width*bands
for idx in range(bands):
imoff = image_offset + idx*lineoff
vrtdsc.AddRawBand(filename, dtype, byteorder,
imoff, pixoff, lineoff)
else:
pixoff=datasize
lineoff=datasize*width
for idx in range(bands):
imoff = image_offset + datasize*width*height*idx
vrtdsc.AddRawBand(filename, dtype, byteorder,
imoff, pixoff, lineoff)
return vrtdsc.GetVRTLines()
def unmix_image(full_image_name, clsf_file, output_name):
# Create the output folder if it does not exist
output_folder = os.path.dirname(output_name)
if not os.path.isdir(output_folder):
os.mkdir(output_folder)
# scale factor (f): 0.5m WV pixel * 1000 = 500m pseudo-MODIS
f = 1000
block_size = f*5 # size of each chunk to load into memory
src_ds = gdal.Open(full_image_name, gdal.GA_ReadOnly)
clsf_ds = gdal.Open(clsf_file, gdal.GA_ReadOnly)
y_dim = src_ds.RasterYSize
x_dim = src_ds.RasterXSize
y_blocks = range(0, y_dim, block_size)
x_blocks = range(0, x_dim, block_size)
# Find the dark point reference
src_dtype = gdal.GetDataTypeSize(src_ds.GetRasterBand(1).DataType)
stretch_params = pp.histogram_threshold(src_ds, src_dtype)
dark_ref = stretch_params[3]
# Calculate the dark point and white point for a
# 2pt correction based on the image histogram
for b in range(len(dark_ref)):
# Add a ceiling to the amount of correction per band (b)
offset = ((b+1)**2)
if dark_ref[b] > 65 - offset and offset < 62:
dark_ref[b] = 65 - offset
white_pt = stretch_params[1]
# Add a const. ceiling to the white correction
if white_pt < 235:
white_pt = 235
print(stretch_params)
# We have to read the entire image ahead of time to calculate the whole image
# melt pond mean, which will be used in each block read below
image_pmean, stage_two_srm = find_image_pmean(src_ds, clsf_ds, dark_ref, white_pt)
# Create an RFC Model from existing training data
model_filename = './rfc_model_worldview.p'
with open(model_filename, 'rb') as mf:
model = pickle.load(mf)
# Flag for setting variables on first iteration
initial = True
pbar = tqdm(total=len(y_blocks)*len(x_blocks), unit='block')
for y in y_blocks:
for x in x_blocks:
read_size_y = check_read_size(y, block_size, y_dim)
read_size_x = check_read_size(x, block_size, x_dim)
if read_size_y != block_size or read_size_x != block_size:
pbar.update()
continue
optic_data = src_ds.ReadAsArray(x, y, read_size_x, read_size_y)
clsf_data = clsf_ds.ReadAsArray(x, y, read_size_x, read_size_y)
# Skip mostly empty blocks
if not valid_block(optic_data):
pbar.update()
continue
print("Rescaling bands...")
# Apply the 2pt correction (simple histogram stretch)
for b in range(7):
optic_data[b, :, :] = pp.rescale_band(optic_data[b, :, :], dark_ref[b], white_pt)
print("Done")
# "Downsample" the classified image. This finds the percentage of each
# pseudo-MODIS pixel that is each surface type.
ds_clsf_data = block_classification(clsf_data, f)
# Analyze the imagery to get relevant data
(srm_list, refl_list, pond_hsv, ocean_hsv,
true_fractions, pond_ocean_diff, pratio) = analyze_imagery(optic_data,
clsf_data, ds_clsf_data,
f, pmean=image_pmean)
# refl_list is a list of all pseudo-MODIS pixels [[b1, b2, ...], [b1, b2, ...], ...]
# srm_list is a list of all the stage 1 reflectance matrices
# Apply stage 1 unmixing
s1_fract, s1_error = stage_one_unmixing(refl_list, srm_list, true_fractions)
# Apply stage 2 unmixing
s2_fract, s2_error = stage_two_unmixing(refl_list, stage_two_srm)
# Unmix using a 'global' srm (generally Rosel's)
global_srm_rosel = np.array([[.08, .16, .95], [.08, .07, .87], [.08, .22, .95], [1, 1, 1]])
# Average of a bunch of WV images, calculated elsewhere
global_srm_wv = np.array([[0.024, 0.201, 0.651], [0.024, 0.152, 0.557], [0.043, 0.279, 0.735], [1, 1, 1]])
s3_fract, s3_error = stage_three_unmixing(refl_list, global_srm_wv)
# Unmix with a machine learning method
s4_fract, s4_error = ml_estimation(refl_list, model)
# Find the difference between the pond color of each individual SRM and the average of all of them
srm_diff = calculate_srm_diff(srm_list, stage_two_srm)
srm_diff2 = calculate_srm_diff(srm_list, global_srm_rosel)
srm_diff3 = calculate_srm_diff(srm_list, global_srm_wv)
# write_to_tds(refl_list, true_fractions)
if initial == True:
initial = False
# Initialize the data output lists
true_fraction_all = true_fractions
s1_fraction_all = s1_fract
s1_error_all = s1_error
s2_fraction_all = s2_fract
s2_error_all = s2_error
s3_fraction_all = s3_fract
s3_error_all = s3_error
s4_fraction_all = s4_fract
s4_error_all = s4_error
pond_hsv_all = pond_hsv
ocean_hsv_all = ocean_hsv
pond_ocean_diff_all = pond_ocean_diff
srm_diff_all = srm_diff
pratio_all = pratio
srm_diff2_all = srm_diff2
srm_diff3_all = srm_diff3
else:
# Append the new data to the master lists
true_fraction_all = np.append(true_fraction_all, true_fractions, axis=0)
s1_fraction_all = np.append(s1_fraction_all, s1_fract, axis=0)
s1_error_all = np.append(s1_error_all, s1_error, axis=0)
s2_fraction_all = np.append(s2_fraction_all, s2_fract, axis=0)
s2_error_all = np.append(s2_error_all, s2_error, axis=0)
s3_fraction_all = np.append(s3_fraction_all, s3_fract, axis=0)
s3_error_all = np.append(s3_error_all, s3_error, axis=0)
s4_fraction_all = np.append(s4_fraction_all, s4_fract, axis=0)
s4_error_all = np.append(s4_error_all, s4_error, axis=0)
pond_hsv_all = np.append(pond_hsv_all, pond_hsv, axis=0)
ocean_hsv_all = np.append(ocean_hsv_all, ocean_hsv, axis=0)
pond_ocean_diff_all = np.append(pond_ocean_diff_all, pond_ocean_diff, axis=0)
srm_diff_all = np.append(srm_diff_all, srm_diff, axis=0)
pratio_all = np.append(pratio_all, pratio, axis=0)
srm_diff2_all = np.append(srm_diff2_all, srm_diff2, axis=0)
srm_diff3_all = np.append(srm_diff3_all, srm_diff3, axis=0)
pbar.update()
pbar.close()
src_ds = None
clsf_ds = None
s1_rmse, s2_rmse, s3_rmse, s4_rmse = calculate_rmse(true_fraction_all, s1_fraction_all,
s2_fraction_all, s3_fraction_all, s4_fraction_all)
write_results(output_name, true_fraction_all,
s1_fraction_all, s1_error_all, s1_rmse,
s2_fraction_all, s2_error_all, s2_rmse,
s3_fraction_all, s3_error_all, s3_rmse,
s4_fraction_all, s4_error_all, s4_rmse,
pond_hsv_all, ocean_hsv_all, pond_ocean_diff_all, srm_diff_all, pratio_all)
print("Stage 1 RMSE: {}".format(np.average(s1_rmse)))
print("Stage 2 RMSE: {}".format(np.average(s2_rmse)))
print("Stage 3 RMSE: {}".format(np.average(s3_rmse)))
print("Stage 4 RMSE: {}".format(np.average(s4_rmse)))
print("Pond differences:")
print("Diff from image mean")
print("{0:0.4f}, {1:0.4f}".format(np.mean(srm_diff_all), np.std(srm_diff_all)))
print("Diff from Rosel")
print("{0:0.4f}, {1:0.4f}".format(np.mean(srm_diff2_all), np.std(srm_diff2_all)))
print("Diff from WV global")
print("{0:0.4f}, {1:0.4f}".format(np.mean(srm_diff3_all), np.std(srm_diff3_all)))
def guess_cb(self, *args):
"""Guess image geometry parameters."""
def correlation(array1, array2):
"""Calculate correlation coefficient of two arrays."""
n_elems = float(array1.shape[0])
M1 = Numeric.add.reduce(array1)
M2 = Numeric.add.reduce(array2)
D1 = Numeric.add.reduce(array1 * array1) - M1 * M1 / n_elems
D2 = Numeric.add.reduce(array2 * array2) - M2 * M2 / n_elems
K = (Numeric.add.reduce(array1 * array2) - M1 * M2 / n_elems) / math.sqrt(D1 * D2)
return K
header = long(self.header_entry.get_text())
width = long(self.width_entry.get_text())
height = long(self.height_entry.get_text())
bands = long(self.bands_entry.get_text())
gdaltype = \
gdal.GetDataTypeByName(self.type_list[self.type_menu.get_history()])
numtype = gdalnumeric.GDALTypeCodeToNumericTypeCode(gdaltype)
depth = gdal.GetDataTypeSize(gdaltype) / 8
filename = self.open_entry.get_text()
if os.path.isfile(filename) == 0:
gvutils.error('Input file '+filename+' does not exist!')
return
filesize = os.stat(filename)[ST_SIZE]
if filesize < header:
gvutils.error('Specified header size larger then file size!')
return
imagesize = (filesize - header) / bands / depth
if width != 0 and height == 0:
height = imagesize / width
elif width == 0 and height != 0:
width = imagesize / height
else:
rawfile = open(filename, 'rb')
longt = 40.0 # maximum possible height/width ratio
cor_coef = 0.0
w = long(math.sqrt(imagesize / longt))
w_max = long(math.sqrt(imagesize * longt))
if (self.swap_menu.get_history() == 0 \
and sys.byteorder == 'little') or \
(self.swap_menu.get_history() == 1 and sys.byteorder == 'big'):
swap = 0
else:
swap = 1
while w < w_max:
if imagesize % w == 0:
scanlinesize = w * depth
h = imagesize / w
rawfile.seek(header + h / 2 * scanlinesize)
buf1 = rawfile.read(scanlinesize)
buf2 = rawfile.read(scanlinesize)
a1 = Numeric.fromstring(buf1, numtype)
a2 = Numeric.fromstring(buf2, numtype)
if swap:
a1.byteswapped()
a2.byteswapped()
try:
tmp = correlation(a1.astype(Numeric.Float32), a2.astype(Numeric.Float32))
except:
# catch 0 division errors
gvutils.error('Unable to guess image geometry!')
return
if tmp > cor_coef:
cor_coef = tmp
width = w
height = h
w += 1
self.width_entry.set_text(str(width))
self.height_entry.set_text(str(height))
def main():
# Set Up Arguments
parser = argparse.ArgumentParser()
parser.add_argument("input_dir",
help='''directory path containing date directories of
images to be processed''')
parser.add_argument("image_type",
type=str,
choices=["srgb", "wv02_ms", "pan"],
help="image type: 'srgb', 'wv02_ms', 'pan'")
parser.add_argument("training_dataset", help="training data file")
parser.add_argument("--training_label",
type=str,
default=None,
help="name of training classification list")
parser.add_argument("-o",
"--output_dir",
type=str,
default="default",
help="directory to place output results.")
parser.add_argument("-v",
"--verbose",
action="store_true",
help="display text information and progress")
parser.add_argument("-c",
"--stretch",
type=str,
choices=["hist", "pansh", "none"],
default='hist',
help='''Apply image correction/stretch to input: \n
hist: Histogram stretch \n
pansh: Orthorectify / Pansharpen for MS WV images \n
none: No correction''')
parser.add_argument("--pgc_script",
type=str,
default=None,
help="Path for the pansharpening script if needed")
parser.add_argument("-t",
"--threads",
type=int,
default=16,
help="Number of subprocesses to start")
# Parse Arguments
args = parser.parse_args()
# System filepath that contains the directories or files for batch processing
user_input = args.input_dir
if os.path.isdir(user_input):
src_dir = user_input
src_file = ''
elif os.path.isfile(user_input):
src_dir, src_file = os.path.split(user_input)
else:
raise IOError('Invalid input')
# Image type, choices are 'srgb', 'pan', or 'wv02_ms'
image_type = args.image_type
# File with the training data
tds_file = args.training_dataset
# Default tds label is the image type
if args.training_label is None:
tds_label = image_type
else:
tds_label = args.training_label
# Default output directory
# (if not provided this gets set when the tasks are created)
dst_dir = args.output_dir
threads = args.threads
verbose = args.verbose
stretch = args.stretch
# Use the given pansh script path, otherwise search for the correct folder
# in the same directory as this script.
if args.pgc_script:
pansh_script_path = args.pgc_script
else:
current_path = os.path.dirname(os.path.realpath(__file__))
pansh_script_path = os.path.join(
os.path.split(current_path)[0], 'imagery_utils')
# For Ames OIB Processing:
# White balance flag (To add as user option in future, presently only used on oib imagery)
if image_type == 'srgb':
assess_quality = True
white_balance = True
else:
assess_quality = False
white_balance = False
# Set a default quality score until this value is calculated
quality_score = 1.
# Prepare a list of images to be processed based on the user input
# list of task objects based on the files in the input directory.
# Each task is an image to process, and has a subtask for each split
# of that image.
task_list = utils.create_task_list(os.path.join(src_dir, src_file),
dst_dir)
for task in task_list:
# ASP: Restrict processing to the frame range
# try:
# frameNum = getFrameNumberFromFilename(file)
# except Exception, e:
# continue
# if (frameNum < args.min_frame) or (frameNum > args.max_frame):
# continue
# Skip this task if it is already marked as complete
if task.is_complete():
continue
# Make the output directory if it doesnt already exist
if not os.path.isdir(task.get_dst_dir()):
os.makedirs(task.get_dst_dir())
# Run Ortho/Pan scripts if necessary
if stretch == 'pansh':
if verbose:
print("Orthorectifying and Pansharpening image...")
full_image_name = os.path.join(task.get_src_dir(), task.get_id())
pansh_filepath = pp.run_pgc_pansharpen(pansh_script_path,
full_image_name,
task.get_dst_dir())
# Set the image name/dir to the pan output name/dir
task.set_src_dir(task.get_dst_dir())
task.change_id(pansh_filepath)
# Open the image dataset with gdal
full_image_name = os.path.join(task.get_src_dir(), task.get_id())
if os.path.isfile(full_image_name):
if verbose:
print("Loading image {}...".format(task.get_id()))
src_ds = gdal.Open(full_image_name, gdal.GA_ReadOnly)
else:
print("File not found: {}".format(full_image_name))
continue
# Read metadata to get image date and keep only the metadata we need
metadata = src_ds.GetMetadata()
image_date = pp.parse_metadata(metadata, image_type)
metadata = [image_type, image_date]
# For processing icebridge imagery:
if image_type == 'srgb':
if image_date <= 150:
tds_label = 'spring'
white_balance = True
else:
tds_label = 'summer'
# Load Training Data
tds = utils.load_tds(tds_file, tds_label, image_type)
# tds = utils.load_tds(tds_file, 'srgb', image_type)
if verbose:
print("Size of training set: {}".format(len(tds[1])))
# Set necessary parameters for reading image 1 block at a time
x_dim = src_ds.RasterXSize
y_dim = src_ds.RasterYSize
desired_block_size = 1600
src_dtype = gdal.GetDataTypeSize(src_ds.GetRasterBand(1).DataType)
# Analyze input image histogram (if applying correction)
if stretch == 'hist':
stretch_params = pp.histogram_threshold(src_ds, src_dtype)
else: # stretch == 'none':
# WV Images are actually 11bit stored in 16bit files
if src_dtype > 12:
src_dtype = 11
stretch_params = [
1, 2**src_dtype - 1,
[2**src_dtype - 1 for _ in range(src_ds.RasterCount)],
[1 for _ in range(src_ds.RasterCount)]
]
# Create a blank output image dataset
# Save the classified image output as a geotiff
fileformat = "GTiff"
image_name_noext = os.path.splitext(task.get_id())[0]
dst_filename = os.path.join(task.get_dst_dir(),
image_name_noext + '_classified.tif')
driver = gdal.GetDriverByName(fileformat)
dst_ds = driver.Create(dst_filename,
xsize=x_dim,
ysize=y_dim,
bands=1,
eType=gdal.GDT_Byte,
options=["TILED=YES", "COMPRESS=LZW"])
# Transfer the metadata from input image
# dst_ds.SetMetadata(src_ds.GetMetadata())
# Transfer the input projection and geotransform if they are different than the default
if src_ds.GetGeoTransform() != (0, 1, 0, 0, 0, 1):
dst_ds.SetGeoTransform(
src_ds.GetGeoTransform()) # sets same geotransform as input
if src_ds.GetProjection() != '':
dst_ds.SetProjection(
src_ds.GetProjection()) # sets same projection as input
# Find the appropriate image block read size
block_size_x, block_size_y = utils.find_blocksize(
x_dim, y_dim, desired_block_size)
if verbose:
print("block size: [{},{}]".format(block_size_x, block_size_y))
# close the source dataset so that it can be loaded by each thread seperately
src_ds = None
lock = RLock()
block_queue, qsize = construct_block_queue(block_size_x, block_size_y,
x_dim, y_dim)
dst_queue = Queue()
# Display a progress bar
if verbose:
try:
from tqdm import tqdm
except ImportError:
print("Install tqdm to display progress bar.")
verbose = False
else:
pbar = tqdm(total=qsize, unit='block')
# Set an empty value for the pixel counter
pixel_counts = [0, 0, 0, 0, 0]
NUMBER_OF_PROCESSES = threads
block_procs = [
Process(target=process_block_queue,
args=(lock, block_queue, dst_queue, full_image_name,
assess_quality, stretch_params, white_balance, tds,
metadata)) for _ in range(NUMBER_OF_PROCESSES)
]
for proc in block_procs:
# Add a stop command to the end of the queue for each of the
# processes started. This will signal for the process to stop.
block_queue.put('STOP')
# Start the process
proc.start()
# Collect data from processes as they complete tasks
finished_threads = 0
while finished_threads < NUMBER_OF_PROCESSES:
if not dst_queue.empty():
val = dst_queue.get()
if val is None:
finished_threads += 1
else:
# Keep only the lowest quality score found
quality_score_block = val[0]
if quality_score_block < quality_score:
quality_score = quality_score_block
# Add the pixel counts to the master list
pixel_counts_block = val[1]
for i in range(len(pixel_counts)):
pixel_counts[i] += pixel_counts_block[i]
# Write image data to output dataset
x = val[2]
y = val[3]
classified_block = val[4]
dst_ds.GetRasterBand(1).WriteArray(classified_block,
xoff=x,
yoff=y)
dst_ds.FlushCache()
# Update the progress bar
if verbose: pbar.update()
# Give the other threads some time to finish their tasks.
else:
time.sleep(10)
# Update the progress bar
if verbose: pbar.update()
# Join all of the processes back together
for proc in block_procs:
proc.join()
# Close dataset and write to disk
dst_ds = None
# Write extra data (total pixel counts and quality score to the database (or csv)
output_csv = os.path.join(task.get_dst_dir(),
image_name_noext + '_md.csv')
with open(output_csv, "w") as csvfile:
writer = csv.writer(csvfile)
# writer.writerow(["Quality Score", "White Ice", "Gray Ice", "Melt Ponds", "Open Water", "Shadow"])
writer.writerow([
"Quality Score", "Thick Ice", "Thin Ice", "Shadow",
"Open Water", "Others"
])
writer.writerow([
quality_score, pixel_counts[0], pixel_counts[1],
pixel_counts[2], pixel_counts[3], pixel_counts[4]
])
thumbnail_result = resize_tif_result(dst_filename)
print('thumbnail_result is produced.')
# Close the progress bar
if verbose:
pbar.close()
print("Finished Processing.")