def main(args):
if not os.path.exists(args.savepath):
os.makedirs(args.savepath)
traffic = Traffic.from_file(args.data)
list_features = ["track_unwrapped", "longitude", "latitude", "altitude"]
nb_samples = len(traffic[0])
nb_features = len(list_features)
algo_clustering = AutoencoderTSNE(
gpu=args.gpu,
model=Autoencoder((nb_samples * nb_features, 32, 8, 2)),
learning_rate=args.learning_rate,
weight_decay=args.learning_rate,
lambda_kl=args.lambda_kl,
nb_iterations=args.nb_iterations,
batch_size=args.batch_size,
algo_clustering=DBSCAN(eps=0.06, min_samples=20),
distance_trajectory="euclidean", # delta_max
savepath=f"{args.savepath}/model.pth",
)
t_tsne = traffic.clustering(
nb_samples=None, # nb_samples,
features=list_features,
clustering=algo_clustering,
transform=MinMaxScaler(feature_range=(-1, 1)),
).fit_predict()
t_tsne.to_parquet(f"{args.savepath}/t_tsne.parquet")
def get_flight(inf, calls=['*'], landing=False, takeoff=False):
flist = []
try:
fdata = Traffic.from_file(inf).clean_invalid()
fdata = fdata.filter().eval()
except:
return flist
for flight in fdata:
if (flight.callsign[0:7] == 'WILDLIF'):
continue
if (flight.callsign[0:7] == 'FOLOWME'):
continue
if (flight.callsign[0:5] == 'RADAR'):
continue
if (flight.callsign[0:7] == 'FIRETEN'):
continue
if (flight.callsign[0:8] == 'DUTYOFIR'):
continue
if (np.nanmax(flight.data['altitude'].values) < 300):
continue
if (flight.callsign[0:3] not in calls):
continue
if (landing):
if (np.nanmean(flight.data['vertical_rate'].values) > 200):
continue
if (takeoff):
if (np.nanmean(flight.data['vertical_rate'].values) < -200):
continue
if (flight.typecode is None):
flist.append(flight)
return flist
def get_flight(inf):
"""Load a series of flights from a file using Xavier's 'traffic' library.
Input:
- inf, the input filename
Returns:
- a list of flights
"""
flist = []
# try:
fdata = Traffic.from_file(inf).query("latitude == latitude")
fdata = fdata.clean_invalid().filter().eval()
# except:
# return flist
for flight in fdata:
pos = flight.callsign.find(CNS.search_call)
if (pos < 0):
continue
f_data = flight.data
f_data = f_data.drop_duplicates('timestamp')
f_data = f_data.drop_duplicates('longitude')
f_data = f_data.drop_duplicates('latitude')
f_data = f_data.query('altitude<10000')
f_data = f_data.dropna()
flight.data = f_data
flist.append(flight)
return flist
def openFile(self, *args, **kwargs) -> None:
options = {
"caption":
"Open file",
"filter": ("Pandas DataFrame (*.pkl);;"
"CSV files (*.csv);;"
"Sqlite3 files (*.db)"),
# "filter": "Data files (*.csv *.pkl)",
"directory":
os.path.expanduser("~"),
}
self.filename = QFileDialog.getOpenFileName(self, **options)[0]
if self.filename == "":
return
self.filename = Path(self.filename)
self._traffic = Traffic.from_file(self.filename)
assert self._traffic is not None
self._tview = self._traffic.sort_values("timestamp")
assert self._tview is not None
self.open_dropdown.setItemText(0, self.filename.name)
self.map_plot.default_plot(self._tview)
self.identifier_select.clear()
for callsign in sorted(cast(Set[str], self._tview.callsigns)):
self.identifier_select.addItem(callsign)
self.set_time_range()
self.set_float_columns()
def loadFlights(start_day, end_day):
filename = getFlightsFilename(start_day, end_day)
flights = Traffic.from_file(filename)
if logLvl >= 1:
print("Loaded " + str(len(flights)) + " flights")
return flights
def loadFlights(start_day, end_day, callsign):
filename = recordings_folder + getFlightsFilename(start_day, end_day,
callsign)
flights = Traffic.from_file(filename)
print("Loaded " + str(len(flights)) + " flights")
return flights
def load_flights():
flights=list()
for file in files:
flights.extend(Traffic.from_file(file))
callsigns=dict()
for flight in flights:
if flight.callsign not in callsigns:
callsigns[flight.callsign]=0
callsigns[flight.callsign]+=1
print(callsigns)
filtered=set()
for c in callsigns:
if callsigns[c]>monthly_flights:
filtered.add(c)
print(filtered)
def pretrained_clust(
traffic_file,
list_features,
algo_clustering,
model,
pretrained_path,
to_pickle,
gpu=0,
):
t = Traffic.from_file(traffic_file)
ae_tsne = AutoencoderTSNE(
gpu=gpu,
model=model,
pretrained_path=pretrained_path,
algo_clustering=algo_clustering,
)
t_c = t.clustering(
nb_samples=None,
features=list_features,
clustering=ae_tsne,
transform=MinMaxScaler(feature_range=(-1, 1)),
).fit_predict()
re, scores = ae_tsne.score_samples()
t_c_re = pd.DataFrame.from_records(
[dict(flight_id=f.flight_id, re=re) for f, re in zip(t_c, re)])
t_c = t_c.merge(t_c_re, on="flight_id")
d = {"flight_id": "nunique", "re": ["mean", "min", "max"]}
if scores is not None:
t_c_scores = pd.DataFrame.from_records([
dict(flight_id=f.flight_id, score=score)
for f, score in zip(t_c, scores)
])
t_c = t_c.merge(t_c_scores, on="flight_id")
d["score"] = ["mean", "min", "max"]
print(t_c.groupby(["cluster"]).agg(d))
t_c.to_pickle(to_pickle)
return ae_tsne, t_c
def icaos_monthly_frequency():
all_icaos=set()
for file in files:
flights=Traffic.from_file(file)
icaos=dict()
for flight in flights:
if flight.icao24 not in icaos:
icaos[flight.icao24]=0
icaos[flight.icao24]+=1
#print(icaos)
filtered=set()
for c in icaos:
if icaos[c]>monthly_flights:
filtered.add((c, icaos[c]))
all_icaos.add(c)
print(file[len(recording_folder)+5:len(recording_folder)+12], filtered)
return all_icaos
def callsign_monthly_frequency():
all_callsigns=set()
for file in files:
flights=Traffic.from_file(file)
callsigns=dict()
for flight in flights:
if flight.callsign not in callsigns:
callsigns[flight.callsign]=0
callsigns[flight.callsign]+=1
#print(callsigns)
filtered=set()
for c in callsigns:
if callsigns[c]>monthly_flights:
filtered.add((c, callsigns[c]))
all_callsigns.add(c)
print(file[len(recording_folder)+5:len(recording_folder)+12], filtered)
return all_callsigns
def get_flight_basic(inf):
flist = []
try:
fdata = Traffic.from_file(inf)
except:
return flist
for flight in fdata:
if (flight.callsign[0:7] == 'WILDLIF'):
continue
if (flight.callsign[0:7] == 'FOLOWME'):
continue
if (flight.callsign[0:5] == 'RADAR'):
continue
if (flight.callsign[0:7] == 'FIRETEN'):
continue
if (flight.callsign[0:8] == 'DUTYOFIR'):
continue
if (flight.typecode is None):
flist.append(flight)
return flist
from traffic.data import opensky
from traffic.core import Traffic
import pandas as pd
"""
t_dcm = opensky.history(
"2020-02-26 00:00",
"2020-02-26 23:59",
callsign="DCM%",
)
"""
t_dcm = Traffic.from_file("test.parquet")
#t_dcm = pd.read_parquet("test.parquet")
print(len(t_dcm))
unique_callsigns = set()
for f in t_dcm:
unique_callsigns.add(f.callsign)
print(len(unique_callsigns))
by_callsign = {}
for c in t_dcm:
if c.callsign in by_callsign:
by_callsign[c.callsign].append(c)
else:
by_callsign[c.callsign] = [c]
for c in by_callsign:
if len(by_callsign[c]) == 1:
for p in processes:
p.join()
# Move on to the next block of times
start_dt = start_dt + timedelta(hours=nummer)
# If we have reached the end of the block then exit
if (start_dt >= end_dt):
break
files = glob(odir + '*.pkl')
files.sort()
first = True
for inf in files:
if first:
fdata = Traffic.from_file(inf).query("latitude == latitude")
fdata = fdata.clean_invalid().filter().eval()
first = False
else:
tdata = Traffic.from_file(inf).query("latitude == latitude")
tdata = tdata.clean_invalid().filter().eval()
fdata = fdata + tdata
print(len(fdata))
figsize = (20, 20)
projparm = ccrs.PlateCarree(central_longitude=(bounds[2] - bounds[0]) / 2)
alt_cut = 500
fig, ax = plt.subplots(1,
1,
from traffic.data import opensky
from traffic.core import Traffic, Flight
import pandas as pd
start = "2020-02-26"
end = "2020-02-26"
filename = start + "_" + end + "_DCM_recordings.parquet"
if __name__ == "__main__":
t_dcm = Traffic.from_file(filename)
print(len(t_dcm.callsigns))
flight = t_dcm[8]
typer = 'loiter'
col = 'cyan'
plt.plot(lons, lats, color=col, linewidth=0.3)
return
outf1 = '/home/proud/Desktop/test.svg'
outf2 = '/home/proud/Desktop/test.png'
count = 0
aclist = {}
for inf in files:
print(inf)
# try:
indata = Traffic.from_file(inf)
for flight in indata:
try:
acft = aircraft[flight.icao24]
typ = acft.typecode.values[0]
if (typ in aclist):
aclist[typ][0] = aclist[typ][0] + 1
else:
aclist[typ] = [1, acft.model.values[0]]
except KeyboardInterrupt:
continue
except Exception as e:
print("OH", e)
continue
continue
proc_flight(flight)
def loadFlights(start_day, end_day):
filename = getFlightsFilename(start_day, end_day)
return Traffic.from_file(filename)
