import utils
import glob
from datetime import datetime, timedelta
CSV_DIRECTORY = "input/csv/"
OUTPUT_DIRECTORY = "output/aggregated"
DATE = "20190124"
MINIMUM_WIDTH = 1
start = datetime(2018, 10, 2)
end = datetime(2018, 10, 3) #end day+1!!
SHOWER = "Draconids2018"
STATION = "BEHUMA"
spectrograms = []
for result in utils.perdelta(start, end, timedelta(minutes=5)):
spectrograms.append("RAD_BEDOUR_"+datetime.strftime(result,"%Y%m%d_%H%M")+"_"+STATION+"_SYS001.png")
aggregated_identifications = {}
csv_files = glob.glob(CSV_DIRECTORY+"*.csv")
for spectrogram in spectrograms:
dt = datetime.strptime(spectrogram[11:24], "%Y%m%d_%H%M")
print("{}".format(dt))
#Step 1: read detection file
detection_files = {}
for csv_file in csv_files:
tmp = utils.read_detection_file_per_spectrogram(csv_file,spectrogram)
if tmp is not None:
detection_files[csv_file] = tmp
#Step 2: run meteor identification algorithm
threshold_image = utils.calculate_threshold_image(detection_files)
if header:
writer.writeheader()
writer.writerows(results)
def calculate_distance(image1, image2):
"""Calculate the distance between two images (intersection / union)"""
intersection = np.sum(image1 & image2)
union = np.sum(image1 | image2)
fraction = float(intersection) / float(union)
return fraction
# generate list of all spectrogram names
spectrograms = []
for result in utils.perdelta(START, END, timedelta(minutes=5)):
spectrograms.append("RAD_BEDOUR_" +
datetime.strftime(result, "%Y%m%d_%H%M") + "_" +
STATION + "_SYS001")
csv_files = glob.glob(CSV_DIRECTORY + "*.csv")
write_output([], header=True) #write only the header
for spectrogram in spectrograms:
output = []
# load reference classifications
spectrogram = spectrogram
dt = datetime.strptime(spectrogram[11:24], "%Y%m%d_%H%M")
print "Loading reference classification " + str(dt)
reference_classification = utils.read_detection_file_per_spectrogram(
REFERENCE_FILE, spectrogram + ".wav", swap_topbottom=True)
if reference_classification is None:
import matplotlib.patches as patches
from PIL import Image
import utils
plt.ioff() # Turn interactive plotting off
PNG_DIRECTORY = "input/png/"
CSV_DIRECTORY = "input/csv/"
OUTPUT_DIRECTORY = "output/"
MASKSIZE = (595, 864)
start = datetime(2016, 8, 19)
end = datetime(2016, 8, 20)
station = "BEHUMA"
spectrograms = []
for result in utils.perdelta(start, end, timedelta(minutes=5)):
spectrograms.append("RAD_BEDOUR_"+datetime.strftime(result,"%Y%m%d_%H%M")+"_"+station+"_SYS001.png")
csv_files = glob.glob(CSV_DIRECTORY+"*.csv")
date_time, identifications, volunteers = [], [], []
for spectrogram in spectrograms:
dt = datetime.strptime(spectrogram[11:24], "%Y%m%d_%H%M")
#Step 1: read detection file
detection_files = {}
for csv_file in csv_files:
tmp = utils.read_detection_file_per_spectrogram(csv_file,spectrogram)
if tmp is not None:
detection_files[csv_file] = tmp
#Step 2: run meteor identification algorithm
threshold_image = utils.calculate_threshold_image(detection_files)
#Step 3: select regions that are above identification threshold
plt.yticks(size=12)
plt.ylim([0,1.2*max(binned.counts)])
plt.xlim([binned.index.min(), binned.index.max()])
plt.grid(axis='x')
ax1.xaxis.set_major_formatter(plt.NullFormatter())
ax1.fill_between(binned.index, 0, binned.counts, alpha=.3, color='blue')
ax1.spines["top"].set_visible(False)
ax1.spines["right"].set_visible(False)
# Plot radiant altitude
if show_radiant_altitude:
JDs = map(utils.toJD,rad_pos.keys())
pos = np.deg2rad(rad_pos.values()).tolist()
retrieve = interpolate.interp1d(JDs, pos, axis=0)
ax1b = ax1.twinx()
for utc in utils.perdelta(min(dt), max(dt), timedelta(minutes=60)):
RA, Dec = retrieve(utils.toJD(utc))
equ_coord = sidereal.RADec(RA,Dec)
h = equ_coord.hourAngle(utc,lon)
horiz_coord = equ_coord.altAz(h,lat)
ax1b.set_ylabel('Radiant elevation [$\degree$]',size=14,style='italic',color='green')
ax1b.scatter(utc,np.rad2deg(horiz_coord.alt),marker='.',color='green')
ax1.autoscale(True,axis='x',tight=True)
ax2 = plt.subplot(2, 1, 2)
ax2.spines["top"].set_visible(False)
ax2.spines["right"].set_visible(False)
col_list = [ utils.color_gradient(val) for val in binned.percentage_completed]
plt.bar(binned.index, height=binned.percentage_completed, width=1/24., color = col_list, ec='black')
locs, labels = plt.xticks()
plt.setp(labels, rotation=45)
plt.yticks(size=12)
plt.ylim([0, 1.2 * max(binned.counts)])
plt.xlim([binned.index.min(), binned.index.max()])
plt.grid(axis='x')
ax1.xaxis.set_major_formatter(plt.NullFormatter())
ax1.fill_between(binned.index, 0, binned.counts, alpha=.3, color='blue')
ax1.spines["top"].set_visible(False)
ax1.spines["right"].set_visible(False)
# Plot radiant altitude
if show_radiant_altitude:
JDs = map(utils.toJD, rad_pos.keys())
pos = np.deg2rad(rad_pos.values()).tolist()
retrieve = interpolate.interp1d(JDs, pos, axis=0)
ax1b = ax1.twinx()
for utc in utils.perdelta(min(dt), max(dt), timedelta(minutes=60)):
RA, Dec = retrieve(utils.toJD(utc))
equ_coord = sidereal.RADec(RA, Dec)
h = equ_coord.hourAngle(utc, lon)
horiz_coord = equ_coord.altAz(h, lat)
ax1b.set_ylabel('Radiant elevation [$\degree$]',
size=14,
style='italic',
color='green')
ax1b.scatter(utc,
np.rad2deg(horiz_coord.alt),
marker='.',
color='green')
ax1.autoscale(True, axis='x', tight=True)
ax2 = plt.subplot(2, 1, 2)
fieldnames = ['spectrogram', 'outlier_threshold', 'cost']
writer = csv.DictWriter(csvfile, fieldnames=fieldnames)
if header:
writer.writeheader()
writer.writerows(results)
def calculate_distance(image1, image2):
"""Calculate the distance between two images (intersection / union)"""
intersection = np.sum(image1 & image2)
union = np.sum(image1 | image2)
fraction = float(intersection) / float(union)
return fraction
# generate list of all spectrogram names
spectrograms = []
for result in utils.perdelta(START, END, timedelta(minutes=5)):
spectrograms.append("RAD_BEDOUR_"+datetime.strftime(result,"%Y%m%d_%H%M")+"_"+STATION+"_SYS001")
csv_files = glob.glob(CSV_DIRECTORY+"*.csv")
write_output([], header=True) #write only the header
for spectrogram in spectrograms:
output = []
# load reference classifications
spectrogram = spectrogram
dt = datetime.strptime(spectrogram[11:24], "%Y%m%d_%H%M")
print "Loading reference classification "+str(dt)
reference_classification = utils.read_detection_file_per_spectrogram(REFERENCE_FILE,spectrogram+".wav", swap_topbottom=True)
if reference_classification is None:
continue
print("--- %s seconds ---" % (time.time() - start_time))
