def multiTemplateMatch(stTempLow, stLow, threshLow, stTempHigh, stHigh,
threshHigh, numComp, tolerance, distance):
# make a couple useful list
detectionsTemp = []
detections = []
# iterate through each channel
for s in range(len(stTempLow)):
# call the template matching function in each band
#detectionsLow,sl = correlation_detector(obspy.Stream(stLow[s]),obspy.Stream(stTempLow[s]),threshLow,tolerance)
#detectionsHigh,sh = correlation_detector(obspy.Stream(stHigh[s]),obspy.Stream(stTempHigh[s]),threshHigh,tolerance)
detectionsLow, sl = correlation_detector(obspy.Stream(stLow[s]),
obspy.Stream(stTempLow[s]),
threshLow, distance)
detectionsHigh, sh = correlation_detector(obspy.Stream(stHigh[s]),
obspy.Stream(stTempHigh[s]),
threshHigh, distance)
#print(len(detectionsLow))
#print(len(detectionsHigh))
# get all high frequency trigger times for today
detHighTimes = []
for i in range(len(detectionsHigh)):
detHighTimes.append(detectionsHigh[i].get("time"))
# loop through all low frequency triggers for today
for i in range(len(detectionsLow)):
detLowTime = detectionsLow[i].get("time")
# calculate time difference between low freq trigger and all high freq triggers
diffs = np.subtract(detLowTime, detHighTimes)
# only interested in positive values of 'diffs', which indicates high freq trigger first
diffs[diffs < -1 * tolerance] = float("nan")
# save low freq trigger if a high freq trigger is sufficiently close
if len(diffs) > 0:
if min(diffs) < tolerance:
detectionsTemp.append(detLowTime)
# sort detections chronologically
detectionsTemp.sort()
#print(detectionsTemp)
# save detections if they show up on desired number of components
if len(detectionsTemp) > 0:
for d in range(len(detectionsTemp) - numComp - 1):
#print(detectionsTemp[d+numComp-1] - detectionsTemp[d])
if detectionsTemp[d + numComp - 1] - detectionsTemp[d] < tolerance:
detections.append(detectionsTemp[d])
return detections
def xcorr(self, data, template, detect_value=0.5):
# pass_step = 0.1 # pass step in seconds
# samples / seconds
# 128 s 1 sec
# 64 s 0.5 sec
# 32 s 0.25 sec +
# 16 s 0.125 sec
detections, sims = correlation_detector(stream=data,
templates=template,
heights=detect_value,
distance=self.corr_step,
plot=None)
if len(detections) != 0:
# exclude self correlation
detections = [v for v in detections if v['similarity'] < 0.999]
# del detections[0]
return detections, sims
def test_correlate_stream_template_and_correlation_detector(self):
template = read().filter('highpass', freq=5).normalize()
pick = UTCDateTime('2009-08-24T00:20:07.73')
template.trim(pick, pick + 10)
n1 = len(template[0])
n2 = 100 * 3600 # 1 hour
dt = template[0].stats.delta
# shift one template Trace
template[1].stats.starttime += 5
stream = template.copy()
np.random.seed(42)
for tr, trt in zip(stream, template):
tr.stats.starttime += 24 * 3600
tr.data = np.random.random(n2) - 0.5 # noise
if tr.stats.channel[-1] == 'Z':
tr.data[n1:2 * n1] += 10 * trt.data
tr.data = tr.data[:-n1]
tr.data[5 * n1:6 * n1] += 100 * trt.data
tr.data[20 * n1:21 * n1] += 2 * trt.data
# make one template trace a bit shorter
template[2].data = template[2].data[:-n1 // 5]
# make two stream traces a bit shorter
stream[0].trim(5, None)
stream[1].trim(1, 20)
# second template
pick2 = stream[0].stats.starttime + 20 * n1 * dt
template2 = stream.slice(pick2 - 5, pick2 + 5)
# test cross correlation
stream_orig = stream.copy()
template_orig = template.copy()
ccs = correlate_stream_template(stream, template)
self.assertEqual(len(ccs), len(stream))
self.assertEqual(stream[1].stats.starttime, ccs[0].stats.starttime)
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test if traces with not matching seed ids are discarded
ccs = correlate_stream_template(stream[:2], template[1:])
self.assertEqual(len(ccs), 1)
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test template_time parameter
ccs1 = correlate_stream_template(stream, template)
template_time = template[0].stats.starttime + 100
ccs2 = correlate_stream_template(stream,
template,
template_time=template_time)
self.assertEqual(len(ccs2), len(ccs1))
delta = ccs2[0].stats.starttime - ccs1[0].stats.starttime
self.assertAlmostEqual(delta, 100)
# test if all three events found
detections, sims = correlation_detector(stream, template, 0.2, 30)
self.assertEqual(len(detections), 3)
dtime = pick + n1 * dt + 24 * 3600
self.assertAlmostEqual(detections[0]['time'], dtime)
self.assertEqual(len(sims), 1)
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test if xcorr stream is suitable for coincidence_trigger
# result should be the same, return values related
ccs = correlate_stream_template(stream, template)
triggers = coincidence_trigger(None,
0.2,
-1,
ccs,
2,
max_trigger_length=30,
details=True)
self.assertEqual(len(triggers), 2)
for d, t in zip(detections[1:], triggers):
self.assertAlmostEqual(np.mean(t['cft_peaks']), d['similarity'])
# test template_magnitudes
detections, _ = correlation_detector(stream,
template,
0.2,
30,
template_magnitudes=1)
self.assertAlmostEqual(detections[1]['amplitude_ratio'], 100, delta=1)
self.assertAlmostEqual(detections[1]['magnitude'],
1 + 8 / 3,
delta=0.01)
self.assertAlmostEqual(detections[2]['amplitude_ratio'], 2, delta=2)
detections, _ = correlation_detector(stream,
template,
0.2,
30,
template_magnitudes=True)
self.assertAlmostEqual(detections[1]['amplitude_ratio'], 100, delta=1)
self.assertNotIn('magnitude', detections[1])
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test template names
detections, _ = correlation_detector(stream,
template,
0.2,
30,
template_names='eq')
self.assertEqual(detections[0]['template_name'], 'eq')
detections, _ = correlation_detector(stream,
template,
0.2,
30,
template_names=['eq'],
plot=True)
self.assertEqual(detections[0]['template_name'], 'eq')
# test similarity parameter with additional constraints
# test details=True
def simf(ccs):
ccmatrix = np.array([tr.data for tr in ccs])
comp_thres = np.sum(ccmatrix > 0.2, axis=0) > 1
similarity = ccs[0].copy()
similarity.data = np.mean(ccmatrix, axis=0) * comp_thres
return similarity
detections, _ = correlation_detector(stream,
template,
0.1,
30,
similarity_func=simf,
details=True)
self.assertEqual(len(detections), 2)
for d in detections:
self.assertAlmostEqual(np.mean(list(d['cc_values'].values())),
d['similarity'])
# test if properties from find_peaks function are returned
detections, sims = correlation_detector(stream,
template,
0.1,
30,
threshold=0.16,
details=True,
similarity_func=simf)
try:
from scipy.signal import find_peaks # noqa
except ImportError:
self.assertEqual(len(detections), 2)
self.assertNotIn('left_threshold', detections[0])
else:
self.assertEqual(len(detections), 1)
self.assertIn('left_threshold', detections[0])
# also check the _find_peaks function
distance = int(round(30 / sims[0].stats.delta))
indices = _find_peaks(sims[0].data, 0.1, distance, distance)
self.assertEqual(len(indices), 2)
# test distance parameter
detections, _ = correlation_detector(stream, template, 0.2, 500)
self.assertEqual(len(detections), 1)
# test more than one template
# just 2 detections for first template, because second template has
# a higher similarity for third detection
templates = (template, template2)
templatetime2 = pick2 - 10
template_times = (template[0].stats.starttime, templatetime2)
detections, _ = correlation_detector(stream,
templates, (0.2, 0.3),
30,
plot=stream,
template_times=template_times,
template_magnitudes=(2, 5))
self.assertGreater(len(detections), 0)
self.assertIn('template_id', detections[0])
detections0 = [d for d in detections if d['template_id'] == 0]
self.assertEqual(len(detections0), 2)
self.assertEqual(len(detections), 3)
self.assertAlmostEqual(detections[2]['similarity'], 1)
self.assertAlmostEqual(detections[2]['magnitude'], 5)
self.assertEqual(detections[2]['time'], templatetime2)
# test if everything is correct if template2 and stream do not have
# any ids in common
templates = (template, template2[2:])
with warnings.catch_warnings():
warnings.simplefilter("ignore")
detections, sims = correlation_detector(
stream[:1],
templates,
0.2,
30,
plot=True,
template_times=templatetime2,
template_magnitudes=2)
detections0 = [d for d in detections if d['template_id'] == 0]
self.assertEqual(len(detections0), 3)
self.assertEqual(len(detections), 3)
self.assertEqual(len(sims), 2)
self.assertIsInstance(sims[0], Trace)
self.assertIs(sims[1], None)
def multiTemplateMatchDeprecated(path, stat, chans, tempLims, freqLow,
threshLow, freqHigh, threshHigh, tolerance):
# internal control over detection plotting
plotting = False
# define cross correlation search parameters
distance = 10
# make a couple useful list
detectionArray = [[], [], []]
allDetections = []
# make vector of all filenames for a single channel
fileMat = []
filePath = "/media/Data/Data/PIG/MSEED/noIR/" + stat + "/" + chans[0] + "/*"
files = glob.glob(filePath)
files.sort()
fileMat.extend(files)
# specify a specific file (for testing)
fileMat = [
"/media/Data/Data/PIG/MSEED/noIR/PIG2/" + chans[0] + "/2012-05-22." +
stat + "." + chans[0] + ".noIR.MSEED"
]
# get templates for low and high frequency bands
stTempLow = makeTemplates(path, stat, "H*", tempLims, freqLow)
stTempHigh = makeTemplates(path, stat, "H*", tempLims, freqHigh)
# loop through all files- we will replace channel string with wildcard as we go
for f in range(len(fileMat)):
timer = time.time()
# get filename from full path
fname = fileMat[f]
# make filename with wildcard channel
fname = fname.replace(chans[0], "H*")
# pull out day string for user output
day = fname.split("/")[9].split(".")[0]
# read files and do basic preprocessing
stRaw = obspy.read(fname)
stRaw.detrend("demean")
stRaw.detrend("linear")
stRaw.taper(max_percentage=0.01, max_length=10.)
# copy the file
stLow = stRaw.copy()
stHigh = stRaw.copy()
# filter the data to each band
stLow.filter("bandpass", freqmin=freqLow[0], freqmax=freqLow[1])
stHigh.filter("bandpass", freqmin=freqHigh[0], freqmax=freqHigh[1])
# call the template matching function in each band
detectionsLow, sl = correlation_detector(stLow, stTempLow, threshLow,
distance)
detectionsHigh, sh = correlation_detector(stHigh, stTempHigh,
threshHigh, distance)
print(detectionsLow)
print(detectionsHigh)
# at this point, channel consolidation is roughly done
# carry out multiband consolidation
# get all high frequency times
detHighTimes = []
for i in range(len(detectionsHigh)):
detHighTimes.append(detectionsHigh[i].get("time"))
for i in range(len(detectionsLow)):
detLowTime = detectionsLow[i].get("time")
# calculate difference between current detection and all high frequency detections
diffs = abs(np.subtract(detLowTime, detHighTimes))
# save low frequency detections if minimum difference is less than threshold time
if any(diffs) and min(diffs) < tolerance:
allDetections.append(detLowTime)
runtime = time.time() - timer
print("Finished detections for " + day + " in " + str(runtime) +
" seconds\n")
# sort detections chronologically
allDetections.sort()
# get indices of redundant detections
removeInd = []
for d in range(len(allDetections) - 1):
if allDetections[d + 1] - allDetections[d] < 60:
removeInd.append(d + 1)
# replace redundant detections with arbitrary placeholder
for r in removeInd:
allDetections[r] = obspy.UTCDateTime(0)
# make final list of detections and fill with all values not equal to placeholder
finalDetections = []
finalDetections[:] = [
x for x in allDetections if x != obspy.UTCDateTime(0)
]
# give the user some output
print(
str(len(finalDetections)) + " detections found over " +
str(len(fileMat[0])) + " days \n")
if plotting:
# load all traces into one stream for plotting
stHigh += stLow
stHigh += stRaw
plotWinLen = tempLimsLow[1] - tempLimsLow[0]
# plot each detection
for d in range(len(finalDetections)):
stHigh.plot(starttime=finalDetections[d],
endtime=finalDetections[d] + plotWinLen,
equal_scale=False)
return finalDetections
def multiTemplateMatch(path, stat, chans, tempLimsLow, freqLow, threshLow,
tempLimsHigh, freqHigh, threshHigh, tolerance):
# internal control over detection plotting
plotting = False
# make a couple useful list
detectionArray = [[], [], []]
allDetections = []
for c in range(len(chans)):
# make empty vector of detections
detections = []
# get templates for low and high frequency bands
stTempLow = makeTemplates(path, stat, chans[c], tempLimsLow, freqLow)
stTempHigh = makeTemplates(path, stat, chans[c], tempLimsHigh,
freqHigh)
#stTempLow.plot()
# define cross correlation search parameters
distance = 10
# make vector of all filenames
fileMat = []
filePath = "/media/Data/Data/PIG/MSEED/noIR/" + stat + "/" + chans[
c] + "/*"
files = glob.glob(filePath)
files.sort()
fileMat.append(files)
# specify a specific file (for testing)
fileMat = [[
"/media/Data/Data/PIG/MSEED/noIR/PIG2/" + chans[c] +
"/2012-05-22." + stat + "." + chans[c] + ".noIR.MSEED"
]]
# loop through all files
for f in range(len(fileMat[0])):
try:
timer = time.time()
# get filename from full path
fname = fileMat[0][f]
# pull out day string for output
day = fname.split("/")[9].split(".")[0]
# read file
stLow = obspy.read(fname)
# basic preprocessing
stLow.detrend("demean")
stLow.detrend("linear")
stLow.taper(max_percentage=0.01, max_length=10.)
# copy the file
stHigh = stLow.copy()
stRaw = stLow.copy()
# filter the data to each band
stLow.filter("bandpass",
freqmin=freqLow[0],
freqmax=freqLow[1])
stHigh.filter("bandpass",
freqmin=freqHigh[0],
freqmax=freqHigh[1])
# call the template matching function in each band
detectionsLow, sl = correlation_detector(
stLow, stTempLow, threshLow, distance)
detectionsHigh, sh = correlation_detector(
stHigh, stTempHigh, threshHigh, distance)
print(len(detectionsLow))
print(len(detectionsHigh))
# extract time values from detections
differences = np.zeros(
(len(detectionsLow), len(detectionsHigh)))
# calculate detection time difference for each pair of detections
for n in range(len(detectionsLow)):
for m in range(len(detectionsHigh)):
differences[n, m] = detectionsLow[n].get(
"time") - detectionsHigh[m].get("time")
# replace pairs where low frequency detection is first with nan
differences[differences < -1 * distance] = np.nan
try:
# find closest low frequency detection for each high frequency detection
minDiffs = np.nanmin(differences, axis=0)
# keep if the time difference is less than user-inputted threshold
detectDiffs = np.where(minDiffs < tolerance, 1, 0)
# fill detection vector with times returned by high frequency template match
for d in range(len(detectDiffs)):
if detectDiffs[d] == 1:
detections.append(detectionsHigh[d].get("time"))
print("Found " + str(len(detections)) + " detections in " +
fname)
except:
print("No detections for " + fname + "\n")
except:
print("Skipping " + fname + "\n")
# fill array with detections from each file
detectionArray[c].append(detections)
# iterate through first two channels
for c in range(len(chans) - 1):
# iterate through list of detections from channel c
for i in range(len(detectionArray[c][0])):
# iterate through list of detections from channel c+1
for j in range(len(detectionArray[c + 1][0])):
# if both detections are closely spaced in time, add to list of all detections
if np.abs(detectionArray[c][0][i] -
detectionArray[c + 1][0][j]) < tolerance:
allDetections.append(detectionArray[c][0][i])
if c == 0:
# iterate through list of detections from channel c+2 if c == 0 (compare first and third channel)
for j in range(len(detectionArray[c + 2][0])):
# if both detections are closely spaced in time, add to list of all detections
if np.abs(detectionArray[c][0][i] -
detectionArray[c + 2][0][j]) < tolerance:
allDetections.append(detectionArray[c + 2][0][j])
runtime = time.time() - timer
print("Finished detections for " + fname + " in " + str(runtime) +
" seconds\n")
# sort detections chronologically
allDetections.sort()
# get indices of redundant detections
removeInd = []
for d in range(len(allDetections) - 1):
if allDetections[d + 1] - allDetections[d] < 60:
removeInd.append(d + 1)
# replace redundant detections with arbitrary placeholder
for r in removeInd:
allDetections[r] = obspy.UTCDateTime(0)
# make final list of detections and fill with all values not equal to placeholder
finalDetections = []
finalDetections[:] = [
x for x in allDetections if x != obspy.UTCDateTime(0)
]
# give the user some output
print(
str(len(finalDetections)) + " detections found over " +
str(len(fileMat[0])) + " days \n")
if plotting:
# load all traces into one stream for plotting
stHigh += stLow
stHigh += stRaw
plotWinLen = tempLimsLow[1] - tempLimsLow[0]
# plot each detection
for d in range(len(finalDetections)):
stHigh.plot(starttime=finalDetections[d],
endtime=finalDetections[d] + plotWinLen,
equal_scale=False)
return finalDetections
def multiTemplateMatch(path,stat,chans,tempLimsLow,freqLow,threshLow,tempLimsHigh,freqHigh,threshHigh):
# make empty vector of detections
detections = []
# get templates for low and high frequency bands
stTempLow = makeTemplates(path,stat,chans[c],tempLimsLow,freqLow)
stTempHigh = makeTemplates(path,stat,chans[c],tempLimsHigh,freqHigh)
# define cross correlation search parameters
distance = 10
tolerance = 10*60
# make vector of all filenames
fileMat = []
filePath = "/media/Data/Data/PIG/MSEED/noIR/" + stat + "/" + chans[c] + "/*"
files = glob.glob(filePath)
files.sort()
fileMat.append(files)
# specify a specific file (for testing)
fileMat = [["/media/Data/Data/PIG/MSEED/noIR/PIG2/" + chans[c] + "/2012-05-09." + stat + "." + chans[c] + ".noIR.MSEED"]]
#"/media/Data/Data/PIG/MSEED/noIR/PIG2/HHZ/2012-08-21.PIG2.HHZ.noIR.MSEED",
#"/media/Data/Data/PIG/MSEED/noIR/PIG2/HHZ/2012-11-10.PIG2.HHZ.noIR.MSEED"]]
# loop through all files
for f in range(len(fileMat[0])):
# get filename from full path
fname = fileMat[0][f]
# pull out day string for output
day = fname.split("/")[9].split(".")[0]
# read file
stLow = obspy.read(fname)
# basic preprocessing
stLow.detrend("demean")
stLow.detrend("linear")
stLow.taper(max_percentage=0.01, max_length=10.)
# copy the file
stHigh = stLow.copy()
stRaw = stLow.copy()
# filter the data to each band
stLow.filter("bandpass",freqmin=freqLow[0],freqmax=freqLow[1])
stHigh.filter("bandpass",freqmin=freqHigh[0],freqmax=freqHigh[1])
# call the template matching function in each band
detectionsLow,sl = correlation_detector(stLow,stTempLow,threshLow,distance)
detectionsHigh,sh = correlation_detector(stHigh,stTempHigh,threshHigh,distance)
# extract time values from detections
differences = np.zeros((len(detectionsLow),len(detectionsHigh)))
# calculate detection time difference for each pair of detections
for n in range(len(detectionsLow)):
for m in range(len(detectionsHigh)):
differences[n,m] = detectionsLow[n].get("time")-detectionsHigh[m].get("time")
# replace pairs where low frequency detection is first with nan
differences[differences < -2] = np.nan
# find closest low frequency detection for each high frequency detection
minDiffs = np.nanmin(differences,axis=0)
# keep if the time difference is less than user-inputted threshold
detectDiffs = np.where(minDiffs < tolerance,1,0)
# fill detection vector with times returned by high frequency template match
for d in range(len(detectDiffs)):
if detectDiffs[d] == 1:
detections.append(detectionsHigh[d].get("time"))
return detections
def test_correlate_stream_template_and_correlation_detector(self):
template = read().filter('highpass', freq=5).normalize()
pick = UTCDateTime('2009-08-24T00:20:07.73')
template.trim(pick, pick + 10)
n1 = len(template[0])
n2 = 100 * 3600 # 1 hour
dt = template[0].stats.delta
# shift one template Trace
template[1].stats.starttime += 5
stream = template.copy()
np.random.seed(42)
for tr, trt in zip(stream, template):
tr.stats.starttime += 24 * 3600
tr.data = np.random.random(n2) - 0.5 # noise
if tr.stats.channel[-1] == 'Z':
tr.data[n1:2*n1] += 10 * trt.data
tr.data = tr.data[:-n1]
tr.data[5*n1:6*n1] += 100 * trt.data
tr.data[20*n1:21*n1] += 2 * trt.data
# make one template trace a bit shorter
template[2].data = template[2].data[:-n1 // 5]
# make two stream traces a bit shorter
stream[0].trim(5, None)
stream[1].trim(1, 20)
# second template
pick2 = stream[0].stats.starttime + 20 * n1 * dt
template2 = stream.slice(pick2 - 5, pick2 + 5)
# test cross correlation
stream_orig = stream.copy()
template_orig = template.copy()
ccs = correlate_stream_template(stream, template)
self.assertEqual(len(ccs), len(stream))
self.assertEqual(stream[1].stats.starttime, ccs[0].stats.starttime)
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test if traces with not matching seed ids are discarded
ccs = correlate_stream_template(stream[:2], template[1:])
self.assertEqual(len(ccs), 1)
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test template_time parameter
ccs1 = correlate_stream_template(stream, template)
template_time = template[0].stats.starttime + 100
ccs2 = correlate_stream_template(stream, template,
template_time=template_time)
self.assertEqual(len(ccs2), len(ccs1))
delta = ccs2[0].stats.starttime - ccs1[0].stats.starttime
self.assertAlmostEqual(delta, 100)
# test if all three events found
detections, sims = correlation_detector(stream, template, 0.2, 30)
self.assertEqual(len(detections), 3)
dtime = pick + n1 * dt + 24 * 3600
self.assertAlmostEqual(detections[0]['time'], dtime)
self.assertEqual(len(sims), 1)
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test if xcorr stream is suitable for coincidence_trigger
# result should be the same, return values related
ccs = correlate_stream_template(stream, template)
triggers = coincidence_trigger(None, 0.2, -1, ccs, 2,
max_trigger_length=30, details=True)
self.assertEqual(len(triggers), 2)
for d, t in zip(detections[1:], triggers):
self.assertAlmostEqual(np.mean(t['cft_peaks']), d['similarity'])
# test template_magnitudes
detections, _ = correlation_detector(stream, template, 0.2, 30,
template_magnitudes=1)
self.assertAlmostEqual(detections[1]['amplitude_ratio'], 100, delta=1)
self.assertAlmostEqual(detections[1]['magnitude'], 1 + 8 / 3,
delta=0.01)
self.assertAlmostEqual(detections[2]['amplitude_ratio'], 2, delta=2)
detections, _ = correlation_detector(stream, template, 0.2, 30,
template_magnitudes=True)
self.assertAlmostEqual(detections[1]['amplitude_ratio'], 100, delta=1)
self.assertNotIn('magnitude', detections[1])
self.assertEqual(stream_orig, stream)
self.assertEqual(template_orig, template)
# test template names
detections, _ = correlation_detector(stream, template, 0.2, 30,
template_names='eq')
self.assertEqual(detections[0]['template_name'], 'eq')
detections, _ = correlation_detector(stream, template, 0.2, 30,
template_names=['eq'], plot=True)
self.assertEqual(detections[0]['template_name'], 'eq')
# test similarity parameter with additional constraints
# test details=True
def simf(ccs):
ccmatrix = np.array([tr.data for tr in ccs])
comp_thres = np.sum(ccmatrix > 0.2, axis=0) > 1
similarity = ccs[0].copy()
similarity.data = np.mean(ccmatrix, axis=0) * comp_thres
return similarity
detections, _ = correlation_detector(stream, template, 0.1, 30,
similarity_func=simf,
details=True)
self.assertEqual(len(detections), 2)
for d in detections:
self.assertAlmostEqual(np.mean(list(d['cc_values'].values())),
d['similarity'])
# test if properties from find_peaks function are returned
detections, sims = correlation_detector(stream, template, 0.1, 30,
threshold=0.16, details=True,
similarity_func=simf)
try:
from scipy.signal import find_peaks # noqa
except ImportError:
self.assertEqual(len(detections), 2)
self.assertNotIn('left_threshold', detections[0])
else:
self.assertEqual(len(detections), 1)
self.assertIn('left_threshold', detections[0])
# also check the _find_peaks function
distance = int(round(30 / sims[0].stats.delta))
indices = _find_peaks(sims[0].data, 0.1, distance, distance)
self.assertEqual(len(indices), 2)
# test distance parameter
detections, _ = correlation_detector(stream, template, 0.2, 500)
self.assertEqual(len(detections), 1)
# test more than one template
# just 2 detections for first template, because second template has
# a higher similarity for third detection
templates = (template, template2)
templatetime2 = pick2 - 10
template_times = (template[0].stats.starttime, templatetime2)
detections, _ = correlation_detector(stream, templates, (0.2, 0.3), 30,
plot=stream,
template_times=template_times,
template_magnitudes=(2, 5))
self.assertGreater(len(detections), 0)
self.assertIn('template_id', detections[0])
detections0 = [d for d in detections if d['template_id'] == 0]
self.assertEqual(len(detections0), 2)
self.assertEqual(len(detections), 3)
self.assertAlmostEqual(detections[2]['similarity'], 1)
self.assertAlmostEqual(detections[2]['magnitude'], 5)
self.assertEqual(detections[2]['time'], templatetime2)
# test if everything is correct if template2 and stream do not have
# any ids in common
templates = (template, template2[2:])
with warnings.catch_warnings():
warnings.simplefilter("ignore")
detections, sims = correlation_detector(
stream[:1], templates, 0.2, 30, plot=True,
template_times=templatetime2, template_magnitudes=2)
detections0 = [d for d in detections if d['template_id'] == 0]
self.assertEqual(len(detections0), 3)
self.assertEqual(len(detections), 3)
self.assertEqual(len(sims), 2)
self.assertIsInstance(sims[0], Trace)
self.assertIs(sims[1], None)
