def scan(self, regfilt, threshold=5.0):
bins = numpy.arange(self.time.min(),
self.time.max() + 1e-4, self.timebin)
xbins = numpy.arange(self.xmin, self.xmax, 25)
ybins = numpy.arange(self.ymin, self.ymax, 25)
region = Likelihood.Region(self.xmin, self.xmax, self.ymin, self.ymax,
40, regfilt, self.expomap, self.eventfile)
ls = Likelihood.Likelihood(self.X, self.Y, region)
# Build the likelihood model
# Bkg model (isotropic)
iso = Likelihood.Isotropic("bkg", 1.0)
m = Likelihood.GlobalModel("likeModel") + iso
knownSources = []
for i, src in enumerate(self.sources):
thisSrc = Likelihood.PointSource("src%i" % (i + 1), src[0], src[1],
self.eventfile, self.hwu)
m += thisSrc
if i >= 10:
log.info("Fixing normalization of source %s" % (i + 1))
thisSrc["src%i_norm" % (i + 1)].fix()
knownSources.append(thisSrc)
ls.setModel(m)
# Minimize the mlogLike to get the background level
like0 = ls.minimize()
figIntegrated = ls.plot()
excessesFound = []
figs = [figIntegrated]
for t1, t2 in zip(bins[:-1], bins[1:]):
sys.stderr.write(".")
idx = (self.time >= t1) & (self.time < t2)
# db = DbScanner(self.X[idx],self.Y[idx])
# db = SAODetect(self.X[idx],self.Y[idx])
# db = cellDetect.CellDetect(self.X[idx],self.Y[idx],200,3)
ls = Likelihood.Likelihood(self.X[idx], self.Y[idx], region)
totCounts = ls.getTotalCounts()
if totCounts < 3:
# No need to perform any fit. Too few counts in this
# interval
log.debug("Skip interval %s - %s, less than 3 counts here" %
(t1, t2))
continue
# Build the likelihood model
# Bkg model (isotropic)
iso = Likelihood.Isotropic("bkg", 0.1)
m = Likelihood.GlobalModel("likeModel") + iso
for i, src in enumerate(knownSources):
m += src
ls.setModel(m)
# Minimize the mlogLike to get the background level
like0 = ls.minimize(verbose=0)
# ls.plot()
bkgLevel = iso['bkg_amp'].value / pow(region.binsize, 2.0)
error = iso['bkg_amp'].error / pow(region.binsize, 2.0)
# print("Bkg: %s +/- %s" %(bkgLevel, error))
db = cellDetect.CellDetect(self.X[idx], self.xmin, self.xmax,
self.Y[idx], self.ymin, self.ymax, 200,
3)
centroids = db.findExcesses(bkgLevel * pow(200, 2),
error * pow(200, 2))
newSources = []
if len(centroids) > 0:
# Verify if this source is truly significant
TSs = []
for (x, y) in centroids:
sys.stderr.write("-")
# Avoid computing the TS for a source too close to the ones already
# known
d = map(
lambda src: math.sqrt(
pow(src.xpos - x, 2) + pow(src.ypos - y, 2)),
knownSources)
dd = filter(lambda x: x < 5.0 / 0.05, d)
if len(dd) > 0:
# There is a source close to this centroid. No need
# to investigate further
continue
thisSrc = Likelihood.PointSource("testsrc", x, y,
self.eventfile, self.hwu,
knownSources[0].outfile)
ls.model += thisSrc
like1 = ls.minimize(verbose=0)
TSs.append(2 * (like0 - like1))
# print("(X,Y) = (%s,%s) -> TS = %s" %(x,y,TSs[-1]))
ls.model.removeSource("testsrc")
# Using the approximation TS = sqrt( significance )
if TSs[-1] >= pow(threshold, 2):
newSources.append([x, y, TSs[-1]])
if len(newSources) > 0:
db.centroids = numpy.array(newSources)
fig = db.plot(xbins, ybins)
figs.append(fig)
for (x, y, ts) in newSources:
excessesFound.append([t1, t2, x, y, math.sqrt(ts)])
sys.stderr.write("\n")
return excessesFound, figs
