def analyze_region(spectrogram: Spectrogram, roi: dict):
"""
"""
time, velocity, intensity = spectrogram.slice(roi['time'], roi['velocity'])
# Do we want to convert to power?
power = spectrogram.power(intensity)
# let's look at the peak power at each time
peaks = np.max(power, axis=0)
speaks = sorted(peaks)
threshold = speaks[int(len(peaks) * 0.04)]
amax = np.argmax(power, axis=0)
power[power < threshold] = 0.0
span = 30
nmax = len(velocity) - 1
times, centers, widths, amps = [], [], [], []
def peg(x):
if x < 0:
return 0
if x > nmax:
return nmax
return x
for t in range(len(time) - 1):
if power[amax[t], t] > threshold:
acenter = amax[t]
vfrom, vto = peg(acenter - span), peg(acenter + span)
gus = Gaussian(velocity[vfrom:vto], power[vfrom:vto, t])
if gus.valid:
times.append(time[t])
centers.append(gus.center)
widths.append(gus.width)
amps.append(gus.amplitude)
else:
print(f"[{t}] {gus.error}")
#power[power>=threshold] = 1
plt.pcolormesh(time * 1e6, velocity, power)
plt.plot(np.array(times) * 1e6, centers, 'r.', alpha=0.5)
plt.show()
def __init__(self, spec: Spectrogram, time_chunk: int = 16, velocity_chunk: int = 16):
self.spectrogram = spec
self.time_chunk = time_chunk
self.velocity_chunk = velocity_chunk
powers = spec.power(spec.intensity)
nrows, ncols = powers.shape
nrows = nrows // velocity_chunk
ncols = ncols // time_chunk
self.intensity = ar = np.zeros((nrows, ncols))
rows = np.linspace(0, nrows * velocity_chunk,
num=nrows + 1, dtype=np.uint32)
cols = np.linspace(0, ncols * time_chunk,
num=ncols + 1, dtype=np.uint32)
for row in range(nrows):
rfrom, rto = rows[row], rows[row + 1]
for col in range(ncols):
ar[row, col] = np.mean(
powers[rfrom:rto, cols[col]:cols[col + 1]])
self.time = spec.time[0:len(spec.time):time_chunk]
self.velocity = spec.velocity[0:len(spec.velocity):velocity_chunk]
vals = ar.flatten()
vals.sort()
self.spower = vals
self.threshold = vals[int(len(vals) * 0.8)]
class SpectrogramWidget:
"""
A Jupyter notebook widget to represent a spectrogram, along with
numerous controls to adjust the appearance of the spectrogram.
For the widget to behave in a Jupyter notebook, place::
%matplotlib widget
at the top of the notebook. This requires that the package
ipympl is installed, which can be done either with pip3
or conda install ipympl.
I also recommend editing ~/.jupyter/custom/custom.css to
modify the definition of .container::
.container {
width: 100% !important;
margin-right: 40px;
margin-left: 40px;
}
**Inputs**
- digfile: either a string or DigFile
- kwargs: optional keyword arguments. These are passed to
the Spectrogram constructor and to the routine that
creates the control widgets.
**Data members**
- digfile: the source data DigFile
- title: the title displayed above the spectrogram
- baselines: a list of baseline velocities
- spectrogram: a Spectrogram object deriving from digfile
- fig:
- axSpectrum:
- axSpectrogram:
- image:
- colorbar:
- individual_controls: dictionary of widgets
- controls:
"""
_gspec = {
'width_ratios': [6, 1.25],
'height_ratios': [1],
'wspace': 0.05,
'left': 0.075,
'right': 0.975,
}
def __init__(self, *args, **kwargs):
"""
If one passes in a single unnamed arg, it can either be a digfile,
a string pointing to a digfile, or a two-dimensional ndarray.
If we are founded on a dig file, it is possible to recompute
things. Only a subset of operations are possible when we're
based on a two-dimensional array, but perhaps that is sometimes
desirable.
"""
self.digfile = None
if len(args) == 1:
arg = args[0]
if isinstance(arg, str):
self.digfile = DigFile(arg)
elif isinstance(arg, DigFile):
self.digfile = arg
if self.digfile == None and len(args):
# Let's see if we have enough information to display a spectrogram
# That means we have a two-dimensional ndarray, and possibly corresponding
# time and velocity arrays. The signature would be (intensity, [times, velocities]).
arg = args[0]
if isinstance(arg, np.ndarray) and len(arg.shape) == 2:
self._static = {
'intensity': arg,
}
if len(args) == 3:
self._static['time'] = np.asarray(args[1])
self._static['velocity'] = np.asarray(args[2])
else:
self._static['time'] = np.arange(0, -1 + arg.shape[1])
self._static['velocity'] = np.arange(0, -1 + arg.shape[0])
assert hasattr(
self, '_static'
), "Inappropriate arguments passed to the SpectrogramWidget constructor"
# If LaTeX is enabled in matplotlib, underscores in the title
# cause problems in displaying the histogram. However, we can
# solve this by not using latex in displaying the title, so there
# is no need to alter characters here.
self.title = "" if self.static else self.digfile.filename.split(
'/')[-1]
self.baselines = []
# Compute the base spectrogram (do we really need this?)
self.spectrogram = None
if self.dig:
self.spectrogram = Spectrogram(self.digfile, None, None, **kwargs)
self.spectrogram_fresh = True # flag for the first pass
self.spectrogram.overlap = 0.875
self.fig, axes = plt.subplots(nrows=1,
ncols=2,
sharey=True,
squeeze=True,
gridspec_kw=self._gspec)
self.axSpectrogram, self.axSpectrum = axes
self.subfig = None
self.axTrack = None
self.axSpare = None
# At the moment, clicking on the image updates the spectrum
# shown on the left axes. It would be nice to be more sophisticated and
# allow for more sophisticated interactions, including the ability
# to display more than one spectrum.
self.fig.canvas.mpl_connect('button_press_event',
lambda x: self.handle_click(x))
self.fig.canvas.mpl_connect('key_press_event',
lambda x: self.handle_key(x))
self.spectrum(None, "")
self.image = None # we will set in update_spectrogram
self.colorbar = None # we will set this on updating, based on the
self.peak_followers = [] # will hold any PeakFollowers
self.spectra = [] # will hold spectra displayed at right
self.spectra_in_db = True # should spectra be displayed in db?
self.controls = dict() # widgets stored by name
self.layout = None # how the controls get laid out
self.selecting = False # we are not currently selecting a ROI
self.roi = [] # and we have no regions of interest
self.threshold = None
self.make_controls(**kwargs)
display(self.layout)
self.update_spectrogram()
@property
def static(self):
"If we are not associated with a dig file, return True"
return hasattr(self, '_static')
@property
def dig(self):
"True if we are associated with a dig file and can recompute the spectrogram"
return isinstance(self.digfile, DigFile)
@property
def intensity(self):
"Return the two-dimensional array of intensities"
if self.dig:
return self.spectrogram.intensity
else:
return self._static['intensity']
def make_controls(self, **kwargs):
"""
Create the controls for this widget and store them in self.controls.
"""
cd = self.controls # the dictionary of controls
t_range = kwargs.get('t_range', (0, 100))
# If we are associated with a dig file, include controls that
# allow recomputation of the spectrogram
if self.dig:
df = self.digfile
# FFT size ###########################################
# Set the size of each spectrum
pps = self.spectrogram.points_per_spectrum
# pps = kwargs.get('points_per_spectrum', 8192)
val = int(np.log2(pps))
cd['spectrum_size'] = slide = widgets.IntSlider(
value=val, min=8, max=18, step=1, description="FFT 2^n")
slide.continuous_update = False
slide.observe(
lambda x: self.overhaul(points_per_spectrum=2**x['new']),
names="value")
# Set the overlap percentage of successive time intervals
cd['overlap'] = slide = widgets.FloatSlider(
description='Overlap %',
value=100.0 * self.spectrogram.overlap,
min=0,
max=100)
slide.continuous_update = False
slide.observe(lambda x: self.overhaul(overlap=x['new'] * 0.01),
names="value")
# Thumbnail ###########################################
# Display a thumbnail of the raw signal
cd['raw_signal'] = widgets.ToggleButton(value=False,
description="Show V(t)",
button_style='',
icon='check')
cd['raw_signal'].observe(lambda b: self.show_raw_signal(b),
names="value")
# Time range ###########################################
cd['t_range'] = slide = ValueSlider("Time (µs)",
t_range, (df.t0, df.t_final),
1e6,
readout_format=".1f")
slide.continuous_update = False
# Velocity range ###########################################
cd['velocity_range'] = slide = ValueSlider(
"Velocity (km/s)", (0, 50), (0.0, self.spectrogram.v_max),
1e-3,
readout_format=".1f",
continuous_update=False)
else:
d = self._static
cd['t_range'] = ValueSlider("Time",
t_range, (d['time'][0], d['time'][-1]),
1e-3,
readout_format=".1f",
continuous_update=False)
cd['velocity_range'] = ValueSlider("Velocity (km/s)", (0, 100),
(0.0, d['velocity'][-1]),
1e-3,
readout_format=".1f",
continuous_update=False)
cd['t_range'].observe(lambda x: self.do_update(x), names="value")
cd['velocity_range'].observe(lambda x: self.update_velocity_range(x),
names="value")
# Color range ###########################################
imax = self.intensity.max()
if self.dig:
hl = self.spectrogram.histogram_levels
imin = hl['tens'][3]
# Let's figure out the range likely to produce a clear
# image. Put the 50% point at the bottom end and the 95%
# at the top?
def scale(x):
return int(100.0 * (x - imin) / (imax - imin))
start_range = (scale(hl['tens'][8]), scale(hl['tenths'][9]))
else:
imin = imax - 150 # ?? this is bad and assumes db
start_range = (40, 70)
cd['intensity_range'] = slide = ValueSlider("Color",
start_range, (imin, imax),
multiplier=1,
readout_format=".0f",
continuous_update=False)
slide.observe(lambda x: self.update_color_range(), names="value")
# Threshold percentage #####################################
cd['threshold'] = slide = widgets.FloatSlider(
description='Noise floor %',
value=0,
min=0,
max=100.0,
continuous_update=False)
slide.observe(lambda x: self.update_threshold(x['new']), names="value")
# Color map selector ###########################################
the_maps = sorted(COLORMAPS.keys())
the_maps.append('Computed')
cd['color_map'] = widgets.Dropdown(
options=the_maps,
value=DEFMAP,
description='Color Map',
disabled=False,
)
cd['color_map'].observe(lambda x: self.update_cmap(), names="value")
# Click selector ###########################################
# What to do when registering a click in the spectrogram
cd['clicker'] = widgets.Select(options=("Spectrum (dB)", "Spectrum",
"Peak", "Gauss",
"Template_Matching"),
value='Spectrum (dB)',
description="Click:",
disabled=False)
cd['marquee'] = mwidgets.RectangleSelector(
self.axSpectrogram,
lambda eclick, erelease: self.RSelect(eclick, erelease),
interactive=True,
useblit=True,
rectprops=dict(facecolor='yellow',
edgecolor='red',
alpha=0.2,
fill=True),
drawtype='box')
cd['marquee'].set_active(False)
# Clear spectra ###########################################
cd['clear_spectra'] = widgets.Button(description="Clear Spectra")
cd['clear_spectra'].on_click(lambda b: self.clear_spectra())
# Clear peak_followers ###########################################
cd['clear_followers'] = widgets.Button(
description="Clear Peak Followers")
cd['clear_followers'].on_click(lambda b: self.clear_followers())
# Computing baselines ###########################################
cd['baselines'] = widgets.Dropdown(options=('_None_', 'Squash', 'FFT'),
value='_None_',
description='Baselines',
disabled=False)
cd['baselines'].observe(lambda x: self.update_baselines(x["new"]),
names="value")
cd['squash'] = widgets.Button(description="Squash in Vertical")
cd['squash'].on_click(lambda b: self.squash_vertical())
columns = [
'color_map;t_range;velocity_range;intensity_range;threshold',
'clicker;clear_spectra;clear_followers',
]
if self.dig:
columns.append('spectrum_size;overlap;raw_signal;baselines;squash')
vboxes = []
for col in columns:
vboxes.append(widgets.VBox([cd[x] for x in col.split(';')]))
self.layout = widgets.HBox(vboxes)
def range(self, var):
"Return the range of the named control, or None if not found."
if var in self.controls:
return self.controls[var].range
return None
def RSelect(self, eclick, erelease):
"Called when self.selecting is True and the marquee is active"
if self.selecting:
t0, t1 = eclick.xdata, erelease.xdata
v0, v1 = eclick.ydata, erelease.ydata
# make sure they are in the right order
if t1 < t0:
t0, t1 = t1, t0
if v1 < v0:
v0, v1 = v1, v0
self.roi.append(dict(time=(t0, t1), velocity=(v0, v1)))
def do_update(self, what):
self.update_spectrogram()
def show_raw_signal(self, box):
"""
Display or remove the thumbnail of the time series data
at the top of the spectrogram window.
"""
if box.new:
# display the thumbnail
t_range = self.range('t_range')
thumb = self.digfile.thumbnail(*t_range)
# we have to superpose the thumbnail on the
# existing velocity axis, so we need to rescale
# the vertical.
tvals = thumb['times'] * 1e6 # convert to µs
yvals = thumb['peak_to_peak']
ylims = self.axSpectrum.get_ylim()
# Map the thumbnail to the top 20%
ymax = yvals.max()
yrange = ymax - yvals.min()
yscale = 0.2 * (ylims[1] - ylims[0]) / yrange
vvals = ylims[1] - yscale * (ymax - yvals)
self.raw = self.axSpectrogram.plot(tvals, vvals, 'r-',
alpha=0.5)[0]
else:
try:
self.axSpectrogram.lines.remove(self.raw)
self.raw = None
self.fig.canvas.draw()
self.fig.canvas.flush_events()
except:
pass
def overhaul(self, **kwargs):
"""
A parameter affecting the base spectrogram has been changed, so
we need to recompute everything.
"""
if self.dig:
if self.spectrogram_fresh:
self.spectrogram_fresh = False
else:
self.spectrogram.set(**kwargs)
self.update_spectrogram()
def update_spectrogram(self):
"""
Recompute and display everything
"""
intense = self.spectrogram.intensity if self.dig else self._static[
'intensity']
# Having recomputed the spectrum, we need to set the yrange
# of the color map slider
cmin = intense.min()
cmax = intense.max()
self.controls['intensity_range'].range = (cmin, cmax)
self.display_spectrogram()
def display_spectrogram(self):
"""
"""
trange = self.range('t_range')
vrange = self.range('velocity_range')
if self.dig:
# extract the requisite portions
times, velocities, intensities = self.spectrogram.slice(
trange, vrange)
else:
d = self._static
times, velocities, intensities = d['time'], d['velocity'], d[
'intensity']
# if we have already displayed an image, remove it
if self.colorbar:
self.colorbar.remove()
self.colorbar = None
if self.image:
self.image.remove()
self.image = None
if self.threshold:
intensities[intensities < self.threshold] = self.threshold
self.image = self.axSpectrogram.pcolormesh(times * 1e6, velocities,
intensities)
self.colorbar = self.fig.colorbar(self.image,
ax=self.axSpectrogram,
fraction=0.08)
self.axSpectrogram.set_title(self.title, usetex=False)
self.axSpectrogram.set_xlabel('Time ($\mu$s)')
self.axSpectrogram.set_xlim(*(np.array(trange) * 1e6))
self.axSpectrogram.set_ylabel('Velocity (m/s)')
self.update_velocity_range()
self.update_color_range()
self.update_cmap()
def update_threshold(self, x):
sg = self.spectrogram
if int(x) == 0:
self.threshold = None
else:
if x < 90:
threshold = sg.histogram_levels['tens'][x // 10]
elif x < 99:
threshold = sg.histogram_levels['ones'][int(x - 90)]
else:
threshold = sg.histogram_levels['tenths'][int(10 * (x - 99))]
self.threshold = self.spectrogram.transform(threshold)
self.display_spectrogram()
def update_cmap(self):
"""
Update the color map used to display the spectrogram
"""
mapname = self.controls['color_map'].value
if mapname == 'Computed':
from UI_Elements.generate_color_map import make_spectrogram_color_map
mapinfo = make_spectrogram_color_map(self.spectrogram, 4, mapname)
maprange = (mapinfo['centroids'][1], mapinfo['centroids'][-2])
self.controls['intensity_range'].value = maprange
self.image.set_cmap(COLORMAPS[mapname])
def update_velocity_range(self, info=None):
"""
Update the displayed velocity range using values obtained
from the 'velocity_range' slider.
"""
if info:
old_vmin, old_vmax = info['old']
vmin, vmax = info['new']
if vmax > old_vmax or vmin < old_vmin:
return self.update_spectrogram()
vmin, vmax = self.range('velocity_range')
self.axSpectrogram.set_ylim(vmin, vmax)
self.axSpectrum.set_ylim(vmin, vmax)
def update_color_range(self):
self.image.set_clim(self.range('intensity_range'))
def handle_click(self, event):
try:
# convert time to seconds
t, v = event.xdata * 1e-6, event.ydata
except:
return 0
if self.selecting:
return 0
# Look up what we should do with the click
action = self.controls['clicker'].value
# print(f"The action I am attempting to do is {action}")
try:
if 'Spectrum' in action:
self.spectrum(t, action)
elif 'Template_Matching' in action:
ti = self.spectrogram._time_to_index(t)
vi = self.spectrogram._velocity_to_index(v)
self.match_templates(ti, vi)
else:
# print("I am handling a click that should be a peak follower")
self.follow(t, v, action)
except Exception as eeps:
pass
def handle_key(self, event):
try:
# convert time to seconds
_, v = event.xdata * 1e-6, event.ydata
except:
pass
char = event.key
if char == 'x':
# remove the all spectra
self.clear_spectra()
if char in ('f', 'b', 'F', 'B'):
# We'd like to go exploring
if not hasattr(self, 'explorer_mark'):
self.explorer_mark = 2
else:
shifts = dict(f=4, F=20, b=-4, B=-40)
self.explorer_mark += shifts[char]
self.gaussian_explorer(self.explorer_mark)
if char in ('m', 'M'):
self.selecting = not self.selecting
self.controls['marquee'].set_active(self.selecting)
if char in "0123456789":
n = int(char)
# self.fan_out(int(char))
self.gauss_out(n)
if char in ('a', 'A') and self.roi:
self.analyze_roi()
def clear_spectra(self):
"""Remove all spectra from axSpectrum and the corresponding
markers from axSpectrogram
"""
for x in self.spectra:
self.axSpectrogram.lines.remove(x['marker'])
self.axSpectrum.lines.remove(x['line'])
self.spectra = []
self.fig.canvas.draw()
self.fig.canvas.flush_events()
def clear_followers(self):
"""Remove all followers"""
for x in self.peak_followers:
self.axSpectrogram.lines.remove(x.line)
self.peak_followers = []
self.fig.canvas.draw()
self.fig.canvas.flush_events()
def follow(self, t, v, action):
"""Attempt to follow the path starting with the clicked
point."""
print(
f"Let's follow something starting at {t, v} using action {action}."
)
if action == "Gauss":
fitter = GaussianFitter(self.spectrogram, (t, v))
self.gauss = fitter
elif action == "Peak":
follower = PeakFollower(self.spectrogram, (t, v))
# self.peak = follower
self.peak_followers.append(follower)
follower.run()
tsec, v = follower.v_of_t
follower.line = self.axSpectrogram.plot(tsec * 1e6,
v,
'r.',
alpha=0.4,
markersize=2)[0]
# print("Create a figure and axes, then call self.gauss.show_fit(axes)")
def gauss_out(self, n: int):
"""
Show center velocity, width, and amplitude for gaussian
fits to the data in follower n.
"""
if n >= len(self.peak_followers):
return 0
WRITEOUT, fnum = False, 0
pf = self.peak_followers[n]
times, centers, widths, amps = [], [], [], []
vind = pf.frame['vi_span'].to_numpy()
tind = pf.frame['t_index'].to_numpy()
sp = self.spectrogram
for j in range(len(tind)):
t = sp.time[tind[j]] * 1e6
vfrom, vto = vind[j]
powers = sp.power(sp.intensity[vfrom:vto, tind[j]])
speeds = sp.velocity[vfrom:vto]
gus = Gaussian(speeds, powers)
if gus.valid:
times.append(t)
centers.append(gus.center)
widths.append(gus.width)
amps.append(gus.amplitude)
if WRITEOUT:
fname = f"{os.path.splitext(self.digfile.filename)[0]}_{fnum:04d}.csv"
gus.write_csv(fname)
fnum += 1
if WRITEOUT:
# write out the times, too
fname = f"{os.path.splitext(self.digfile.filename)[0]}_t.csv"
v = np.asarray(times)
np.savetxt(v, fname)
fig, axes = plt.subplots(nrows=1, ncols=3, squeeze=True)
ax1, ax2, ax3 = axes
ax1.errorbar(times, centers, fmt='b-', yerr=widths)
ax1.set_xlabel(r'$t$ ($\mu$s)')
ax1.set_ylabel(r'$v$ (m/s)')
ax2.plot(times, widths, 'r-')
ax2.set_xlabel(r'$t$ ($\mu$s)')
ax2.set_ylabel(r'$\delta v$ (m/s)')
ax3.plot(times, amps, 'g-')
ax3.set_xlabel(r'$t$ ($\mu$s)')
ax3.set_ylabel('Amplitude')
# Store the values for later access
if not hasattr(self, "gauss_outs"):
self.gauss_outs = [None for x in range(len(self.peak_followers))]
else:
while len(self.gauss_outs) < len(self.peak_followers):
self.gauss_outs.append(None)
self.gauss_outs[n] = dict(time=np.array(times),
center=np.array(centers),
width=np.array(widths),
amplitude=np.array(amps))
def gaussian_explorer(self, follower_pt: int):
"""
Show center velocity, width, and amplitude for gaussian
fits to the data in follower n.
"""
if len(self.peak_followers) == 0:
return 0
pf = self.peak_followers[0]
res = pf.results
points = len(res['t_index'])
def bound(x):
return x % points
follower_pt = bound(follower_pt)
# We'd like to show data for this index, the previous one,
# and the next one, along with the gaussian fit
hoods = [pf.hood(n=bound(x + follower_pt)) for x in (-2, -1, 0, 1, 2)]
# Check that we have the requisite figure, and make it if we don't
if not hasattr(self, 'explorer_fig'):
self.explorer_fig, self.explorer_axes = plt.subplots(1,
5,
sharey=True)
# also add a marker to the follower's representation on the
# spectrogram to make it easier to see where we are
self.explorer_marker = self.axSpectrogram.plot([], [], 'k*')[0]
# show where we are
tsec, v = pf.v_of_t
self.explorer_marker.set_data([
tsec[follower_pt] * 1e6,
], [
v[follower_pt],
])
min_v, max_v, max_peak = 1e10, 0, 0
for ax, hood in zip(self.explorer_axes, hoods):
ax.clear()
# plot the data
ax.plot(hood.velocity, hood.intensity, 'ko', alpha=0.5)
# plot the background level used for the moment calculation
bg = hood.moment['background']
ax.plot([hood.velocity[0], hood.velocity[-1]], [bg, bg], 'r-')
# show the center and widths from the moment calculation
tallest = np.max(hood.intensity)
max_peak = max(tallest, max_peak)
ax.plot([
hood.moment['center'] + x * hood.moment['std_err']
for x in (-1, 0, 1)
], 0.5 * tallest * np.ones(3), 'r.')
# show the gaussian
hood.plot_gaussian(ax)
vcenter, width = hood.peak_v, hood.moment['std_dev']
min_v = min(min_v, vcenter - 12 * width)
max_v = max(max_v, vcenter + 12 * width)
# ax.set_xlim(vcenter - 12 * width, vcenter + 12 * width)
ax.set_xlabel(f"$v$ (m/s)")
ax.set_title(f"{hood.time*1e6:.2f}" + " $\\mu$s")
txt = f"m = {hood.moment['center']:.1f} ± {hood.moment['std_err']:.1f}"
txt = f"{txt}\ng = {hood.gaussian.center:.1f} ± {hood.gaussian.width:.1f}"
ax.annotate(txt,
xy=(0.05, 0.95),
xycoords='axes fraction',
horizontalalignment='left',
verticalalignment='top')
for ax in self.explorer_axes:
ax.set_xlim(min_v, max_v)
# label the common velocity axis
ax = self.explorer_axes[0]
ax.set_ylim(-0.05 * max_peak, 1.2 * max_peak)
ax.set_ylabel("Intensity")
def analyze_roi(self):
"""
Extract the region(s) of interest and process them
"""
for roi in self.roi:
analyze_region(self.spectrogram, roi['time'])
def fan_out(self, n: int):
"""Produce a zoomed in version of this trace, showing
the neighborhood around the determined peak.
"""
if n >= len(self.peak_followers):
return 0
pf = self.peak_followers[n]
vind = pf.frame['vi_span'].to_numpy()
tind = pf.frame['t_index'].to_numpy()
self.subfig, axes = plt.subplots(nrows=1,
ncols=2,
sharey=True,
squeeze=True,
gridspec_kw=self._gspec)
self.axSpare, self.axTrack = axes
# We will create a "waterfall" of curves surrounding
# the peaks, each offset by a bit. The x axis will
# represent intensity, with subsequent time traces offset
# by an amount I need to determine. The y axis
# is velocity.
spans = []
vvec = self.spectrogram.velocity # shortcut to velocity vector
tvec = self.spectrogram.time
ivec = self.spectrogram.intensity
# pre-extract a bunch of one-dimensional curves
# and be sure to convert to power
for n in range(len(tind)):
vfrom, vto = vind[n]
spans.append({
'v':
vvec[vfrom:vto],
'power':
self.spectrogram.power(ivec[vfrom:vto, tind[n]]),
't':
tvec[tind[n]] * 1e6,
})
maxima = np.array([np.max(x['power']) for x in spans])
maxpower = maxima.max()
# Let's set the offset between times to be one tenth of
# the maxpower
offset = 0.025 * maxpower
for n in reversed(list(range(len(spans)))):
span = spans[n]
self.axTrack.plot(span['power'] + n * offset,
span['v'],
'b-',
alpha=0.33)
self.axTrack.set_ylabel('$v$')
def squash_vertical(self):
normy = self.spectrogram.squash(dB=False) * 2000
self.axSpectrogram.plot(self.spectrogram.time * 1e6,
normy,
'b-',
alpha=0.75)
def update_baselines(self, method):
"""
Handle the baselines popup menu
"""
from ProcessingAlgorithms.SignalExtraction.baselines import baselines_by_squash
blines = []
self.baselines = [] # remove any existing baselines
if method == "Squash":
peaks, sigs, heights = baselines_by_squash(self.spectrogram)
blines.extend(peaks)
# for n in range(len(heights)):
# if heights[n] > 0.1:
# blines.append(peaks[n])
# Now show the baselines in blines or remove any
# if blines is empty
if not blines:
for b in self.baselines:
self.axSpectrum.lines.remove(b['line'])
self.baselines = [] # remove them
else:
edges = (self.spectrogram.intensity.min(),
self.spectrogram.intensity.max())
for v in blines:
bline = self.axSpectrum.plot([edges[0], edges[1]], [v, v],
'k-',
alpha=0.4)
self.baselines.append(dict(v=v, line=bline))
def baseline_intensity(self):
self.update_baselines("Squash")
figgy = plt.figure()
ax = figgy.add_subplot(1, 1, 1)
sg = self.spectrogram
for bline in self.baselines:
index = sg._velocity_to_index(bline['v'])
if index > 2:
ax.semilogy(sg.time * 1e6, sg.power(sg.intensity[index, :]))
ax.set_xlabel(r"$t$ ($\mu$ s)")
ax.set_ylabel(r"Power")
def Spectrum(self, the_time: float):
"""
Return the column from the spectrogram in power form
"""
sg = self.spectrogram
t_index = sg._time_to_index(the_time)
vals = sg.intensity[:, t_index]
return sg.power(vals)
def spectrum(self, the_time: float, form: str):
"""
Display a spectrum in the left axes corresponding to the
passed value of the_time (which is in seconds).
"""
_colors = ["r", "g", "b", "y"]
if the_time is None:
# Initialize the axes
# self.axSpectrum.plot([0, 1], [0, 1], 'r-')
self.axSpectrum.grid(axis='x', which='both', color='b', alpha=0.4)
else:
if True:
delta_t = self.spectrogram.points_per_spectrum / 2 * \
self.digfile.dt
the_spectrum = Spectrum(self.digfile.values(
the_time - delta_t, the_time + delta_t),
self.digfile.dt,
remove_dc=True)
# compute the level of the 90th percentile
spec = dict(spectrum=the_spectrum)
vals = the_spectrum.db
ordering = np.argsort(vals)
if self.baselines:
blines = [x['v'] for x in self.baselines]
n = -1
while the_spectrum.velocities[ordering[n]] in blines:
n -= 1
else:
n = -1
spec['max'] = vals[ordering[n]]
noise_floor = int(n - 0.1 * len(vals))
spec['90'] = vals[ordering[noise_floor]]
else:
t_index = self.spectrogram._time_to_index(the_time)
vals = self.spectrogram.intensity[:, t_index]
# We need to worry about the format of the spectrum
db = ('dB' in form)
field = 'db' if db else 'power'
the_line = self.axSpectrum.plot(getattr(the_spectrum, field),
the_spectrum.velocities,
_colors[len(self.spectra)],
alpha=0.33)
spec['line'] = the_line[0]
tval = the_time * 1e6 # convert to microseconds
marker = self.axSpectrogram.plot([tval, tval],
[0, self.spectrogram.v_max],
_colors[len(self.spectra)],
alpha=0.33)
spec['marker'] = marker[0]
self.spectra.append(spec)
if db != self.spectra_in_db:
self.spectra_in_db = db # switch our mode
# and replot all the spectra
for spec in self.spectra:
li = spec['line']
sp = spec['spectrum']
li.set(xdata=getattr(sp, field), ydata=sp.velocities)
self.axSpectrum.set_xlabel("Power (dB)" if db else "Power")
if db:
# we should order the values and set a limit at something
# like the strongest decile
ninety = max([x['90'] for x in self.spectra])
peak = max([x['max'] for x in self.spectra])
self.axSpectrum.set_xlim(ninety, peak)
return 0
line = self.axSpectrum.lines[0]
intensities = the_spectrum.db
line.set(xdata=intensities, ydata=the_spectrum.velocities)
# We should also add a line to the spectrogram showing where
# the spectrum came from.
if not self.axSpectrogram.lines:
self.axSpectrogram.plot([0, 0], [0, 1], 'r-', alpha=0.33)
# this won't scale when we add baselines
line = self.axSpectrogram.lines[0]
line.set(xdata=[tval, tval], ydata=[0, self.spectrogram.v_max])
def match_templates(self, time, velocity):
methodsToUse = [
'cv2.TM_CCORR_NORMED', 'cv2.TM_SQDIFF', 'cv2.TM_SQDIFF_NORMED'
]
matcher = TemplateMatcher(
self.spectrogram,
template=Templates.opencv_long_start_pattern5.value,
span=200,
k=20,
methods=methodsToUse)
times, velos, scores, methodsUsed = matcher.match()
time_offset = 2.0 + (2 * abs(
matcher.spectrogram.time[matcher.template_time_offset_index] * 1e6)
)
times = np.array(times) + time_offset
matcher.add_to_plot(self.axSpectrogram,
times,
velos,
scores,
methodsUsed,
show_points=True,
show_medoids=True,
verbose=False,
visualize_opacity=False,
show_bounds=True)
def baselines_by_squash(
spectrogram: Spectrogram,
segments = 8,
min_height = 0.0005, # minimum fraction of the tallest peak for consideration
min_separation = 400, # minimum separation of peaks in m/s
peak_width = 20,
min_percent = 80,
max_side_lobe = 0.5
):
"""
Return a list of baseline velocities and their uncertainties.
Inputs:
- spectrogram: an instance of Spectrogram
Optional inputs:
- segments: an integer specifying the number of segments into which to
divide the time dimension to look for a consistent peak
- min_height: the minimum fraction of the strongest squashed peak
to consider when looking for baselines
- min_separation: the minimum separation of baselines, in m/s
- peak_width: to be considered a peak, look at points at this remove
(in m/s) from a candidate peak to see that their values are lower than
max_side_lobe times the peak's value
- min_percent: minimum percentage of segments satisfying the peak criterion
described above under peak_width
- max_side_lobe: the segment's squashed value at ± peak_width must be
less than or equal to max_side_lobe times the peak's value
Outputs:
- peaks: an array of peaks, in descending order of strength
- widths: an array of uncertainty values for the peaks
- heights: corresponding peak heights, normalized to
the greatest height
"""
# Collapse along the time axis, making sure to use power,
# not dB
dv = spectrogram.velocity[1] - spectrogram.velocity[0]
powers = spectrogram.power(spectrogram.intensity)
# We'd like to squash into about 8 distinct segments
# along axis 1 and then make sure that we get consistent
# values before calling a peak a baseline
boundaries = list(range(0, len(spectrogram.time),
len(spectrogram.time) // segments))
boundaries.append(len(spectrogram.time))
squash = []
for n in range(segments):
squash.append(
powers[:, boundaries[n]:boundaries[n + 1]].mean(axis=1))
combined_spectrum = spectrogram.power(
spectrogram.intensity).mean(axis=1)
tallest = combined_spectrum.max()
# this should be improved
peaks, properties = find_peaks(
combined_spectrum,
height=min_height * tallest, # peaks must be this tall to count
distance=min_separation / dv # peaks must be separated by this much at minimum
)
heights = properties['peak_heights']
peak_ht = heights.max()
# produce an ordering from tallest to shortest peak
ordering = np.flip(np.argsort(heights))
peak_positions = peaks[ordering]
peaks = spectrogram.velocity[peak_positions]
heights = heights[ordering] / peak_ht
# Now we want to filter out the peaks that don't correspond to
# consistent signal across the segments
strong_peaks = []
delta = int(peak_width / dv)
for n, pk_pos in enumerate(peak_positions):
if pk_pos >= delta:
totes = 0
for sq in squash:
try:
if sq[pk_pos + delta] < max_side_lobe * sq[pk_pos] > sq[pk_pos - delta]:
totes += 1
except IndexError:
pass
strong_peaks.append(totes >= segments * min_percent * 0.01)
# print(f"[{pk_pos}] -> {spectrogram.velocity[pk_pos]} ({heights[n]})) got {totes}")
else:
strong_peaks.append(False)
keepers = np.array(strong_peaks)
return peaks[keepers], (np.ones(len(peaks)) * dv)[keepers], heights[keepers]
