Unit 23 - Spatio-temporal parallelization¶
This unit is focused on improving the script created in Unit 22 - Spatio-temporal scripting by processing Sentinel scenes in parallel. Paralelization is performed by ParallelModuleQueue similarly as in Unit 19 - DTM script parallelization, see line 207.
New PyGRASS functionality is introduced on line 136. By
MultiModule it is possible to define a list of
modules which will work as isolated units not influenced by other
processes running parallel. By setting set_temp_region computation
region settings will be also isolated from other processes.
The option ncproc on line 43 enables controlling number of processes running in parallel.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 | #!/usr/bin/env python3
#
##############################################################################
#
# MODULE: ndvi-tgrass-v2
#
# AUTHOR(S): martin
#
# PURPOSE: NDVI TGRASS version 2
#
# DATE: Sat Feb 3 15:45:35 2018
#
##############################################################################
#%module
#% description: NDVI TGRASS script version 2
#%end
#%option G_OPT_STRDS_INPUT
#% key: b4
#% description: Name of the input 4th band space time raster dataset
#%end
#%option G_OPT_STRDS_INPUT
#% key: b8
#% description: Name of the input 8th band space time raster dataset
#%end
#%option G_OPT_STRDS_INPUT
#% key: mask
#% description: Name of the input mask space time raster dataset
#%end
#%option G_OPT_F_OUTPUT
#%end
#%option
#% key: basename
#% description: Basename for output raster maps
#% required: yes
#%end
#%option
#% key: threshold
#% description: Threshold for removing small areas
#% answer: 1600
#%end
#%option
#% key: nprocs
#% description: Number of processes
#% answer: 1
#% type: integer
#%end
import sys
import os
import atexit
from grass.pygrass.modules import Module, MultiModule, ParallelModuleQueue
from grass.script import parser
from grass.script.vector import vector_db_select
def cleanup(idx):
Module('g.remove', flags='f', name='mask' + idx, type='raster')
Module('g.remove', flags='f', name='ndvi' + idx, type='raster')
Module('g.remove', flags='f', name='ndvi_class' + idx, type='raster')
Module('g.remove', flags='f', name='ndvi_class' + idx, type='vector')
def compute(b4, b8, msk, output, idx):
modules = []
modules.append(
Module("g.region",
overwrite = True,
raster = msk,
align = b4,
run_ = False)
)
modules.append(
Module("r.mapcalc",
overwrite = True,
expression = "ndvi{idx} = if(isnull({clouds}), null(), float({b8} - {b4}) / ({b8} + {b4}))".format(
idx=idx, clouds=msk, b8=b8, b4=b4),
run_ = False)
)
recode_str="""-1:0.1:1
0.1:0.5:2
0.5:1:3"""
modules.append(
Module("r.recode",
overwrite = True,
input = "ndvi" + idx,
output = "ndvi_class" + idx,
rules = "-",
stdin_ = recode_str,
run_ = False)
)
colors_str="""1 grey
2 255 255 0
3 green"""
modules.append(
Module("r.colors",
map = "ndvi_class" + idx,
rules = "-",
stdin_ = colors_str,
run_ = False)
)
modules.append(
Module("r.to.vect",
flags = 'sv',
overwrite = True,
input = "ndvi_class" + idx,
output = "ndvi_class" + idx,
type = "area",
run_ = False)
)
modules.append(
Module("v.clean",
overwrite = True,
input = "ndvi_class" + idx,
output = output,
tool = "rmarea",
threshold = options['threshold'],
run_ = False)
)
modules.append(
Module("v.colors",
map=output,
layer="1",
use="cat",
raster="ndvi_class")
)
modules.append(
Module('v.rast.stats',
flags='c',
map=output,
raster='ndvi'+idx,
column_prefix='ndvi',
method=['minimum','maximum','average'],
run_ = False)
)
queue.put(MultiModule(modules, sync=False, set_temp_region=True))
def stats(output, date, fd):
fd.write('-' * 80)
fd.write('\n')
fd.write('NDVI class statistics ({0}: {1})'.format(output, date))
fd.write('\n')
fd.write('-' * 80)
fd.write('\n')
from subprocess import PIPE
ret = Module('v.report', map=output, option='area',
stdout_=PIPE)
for line in ret.outputs.stdout.splitlines()[1:]: # skip first line (cat|label|area)
# parse line (eg. 1||2712850)
data = line.split('|')
cat = data[0]
area = float(data[-1])
fd.write('NDVI class {0}: {1:.1f} ha'.format(cat, area/1e4))
fd.write('\n')
data = vector_db_select(output)
for vals in data['values'].values():
# unfortunately we need to cast values by float
fd.write('NDVI class {0}: {1:.4f} (min) {2:.4f} (max) {3:.4f} (mean)'.format(
vals[0], float(vals[2]), float(vals[3]), float(vals[4])))
fd.write('\n')
def main():
import grass.temporal as tgis
tgis.init()
sp4 = tgis.open_old_stds(options['b4'], 'raster')
sp8 = tgis.open_old_stds(options['b8'], 'raster')
msk = tgis.open_old_stds(options['mask'], 'raster')
idx = 1
data = []
for item in sp4.get_registered_maps(columns='name,start_time'):
b4 = item[0]
date=item[1]
b8 = sp8.get_registered_maps(columns='name',
where="start_time = '{}'".format(date))[0][0]
ms = msk.get_registered_maps(columns='name',
where="start_time = '{}'".format(date))[0][0]
output = '{}_{}'.format(options['basename'], idx)
compute(b4, b8, ms, output, str(idx))
data.append(
(output, date)
)
idx += 1
queue.wait()
idx = 1
fd = open(options['output'], 'w')
for output, date in data:
stats(output, date, fd)
cleanup(str(idx))
idx += 1
fd.close()
return 0
if __name__ == "__main__":
options, flags = parser()
# queue for parallel jobs
queue = ParallelModuleQueue(int(options['nprocs']))
sys.exit(main())
|
Sample script to download: ndvi-tgrass-v2.py
Úkol
Compare ndvi-tgrass-v1.py / ndvi-tgrass-v2.py processing time