#!/usr/bin/local/python
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.patches as plt_patches
from pypad import cspad
from pypad import utils
from pypad import read
from pypad import plot
cs1 = cspad.CSPad.load("examples/my_cspad.cspad")
cs2 = cspad.CSPad.default()
cs3 = cspad.CSPad.default()
cs3.quad_rotation[0] = 10.
cs3.quad_rotation[1] = 0.
cs3.quad_rotation[2] = 2.
cs3.quad_rotation[3] = -10.
raw_image = read.load_raw_image("examples/gold-minus490mm.h5")
fig = plt.figure()
ax = plt.subplot(121)
plot.imshow_cspad(cs3(raw_image), ax=ax, scrollable=True)
ax = plt.subplot(122)
plot.sketch_2x1s(cs3.pixel_positions, ax)
plt.show()
def _objective(self, param_vals, raw_image):
"""
The objective function for finding a good center. Minimize this.
Parameters
----------
param_vals : ndarray
A flat array of the parameters we're optimizing. Which parameters
these are depends on the entires in `list_of_params`.
list_of_params : list of str
A list of which parameters are contained in `params_to_opt`.
Returns
-------
obj : float
The value of the function. Lower is better.
See Also
--------
self._inject_params_into_dict : function
Makes sense of `params_to_opt` and `list_of_params`
"""
# un-ravel & inject the param values in the geometry object
param_dict = self._unravel_params(param_vals)
self.cspad.set_many_params(param_dict.keys(), param_dict.values())
# compute the radial profile
bc, bv = self.cspad.intensity_profile(raw_image, n_bins=None)
if self.radius_range == None:
bin_centers, bin_values = bc, bv
else:
bin_centers, bin_values = self._slice(bc, bv)
# if plotting is requested, plot away!
if self.plot_each_iteration:
self._axL.cla()
self._axR.cla()
plot.sketch_2x1s(self.cspad.pixel_positions, self._axL)
self._axR.plot(bin_centers, bin_values, lw=2, color='k')
self._axR.set_xlabel('Radius (mm)')
self._axR.set_ylabel('Intensity (arb. units)')
self._axR.set_ylim((0, bin_values.max() * 1.1))
plt.draw()
# --------- HERE IS THE OBJECTIVE FUNCTION -- MODIFY TO PLAY -----------
# test to see if any ASICS are overlapping, and if they are return a big
# number so the optimizer avoids those regions
if self.cspad.do_asics_overlap:
print "Move caused ASIC overlap: rejecting it"
return 1.0e300
# new objective function : overlap integral
if self.objective_type == 'overlap':
bins = len(bin_centers)
quad_profiles = []
for i in range(4):
bc, bv = self.cspad.intensity_profile(raw_image,
n_bins=bins,
quad=i)
quad_profiles.append(bv)
m = np.min([a.shape for a in quad_profiles])
quad_profiles = np.array([a[:m] for a in quad_profiles])
obj = -np.sum(np.product(quad_profiles, axis=0))
# old objective function : peak height
elif self.objective_type == 'peak_height':
obj = -bin_values.max()
if self.width_weight != 0.0:
obj += self.width_weight * self._simple_width(
bin_values, bar=self.horizontal_cut)
if self.peak_weight != 0.0:
obj += self.peak_weight
else:
raise ValueError('No implemented objective_type: %s' %
objective_type)
# ----------------------------------------------------------------------
print "objective value: %.4e" % obj
return obj
def _objective(self, param_vals, raw_image):
"""
The objective function for finding a good center. Minimize this.
Parameters
----------
param_vals : ndarray
A flat array of the parameters we're optimizing. Which parameters
these are depends on the entires in `list_of_params`.
list_of_params : list of str
A list of which parameters are contained in `params_to_opt`.
Returns
-------
obj : float
The value of the function. Lower is better.
See Also
--------
self._inject_params_into_dict : function
Makes sense of `params_to_opt` and `list_of_params`
"""
# un-ravel & inject the param values in the geometry object
param_dict = self._unravel_params(param_vals)
self.cspad.set_many_params(param_dict.keys(), param_dict.values())
# compute the radial profile
bc, bv = self.cspad.intensity_profile(raw_image, n_bins=None)
if self.radius_range == None:
bin_centers, bin_values = bc, bv
else:
bin_centers, bin_values = self._slice(bc, bv)
# if plotting is requested, plot away!
if self.plot_each_iteration:
self._axL.cla()
self._axR.cla()
plot.sketch_2x1s(self.cspad.pixel_positions, self._axL)
self._axR.plot(bin_centers, bin_values, lw=2, color='k')
self._axR.set_xlabel('Radius (mm)')
self._axR.set_ylabel('Intensity (arb. units)')
self._axR.set_ylim((0, bin_values.max()*1.1))
plt.draw()
# --------- HERE IS THE OBJECTIVE FUNCTION -- MODIFY TO PLAY -----------
# test to see if any ASICS are overlapping, and if they are return a big
# number so the optimizer avoids those regions
if self.cspad.do_asics_overlap:
print "Move caused ASIC overlap: rejecting it"
return 1.0e300
# new objective function : overlap integral
if self.objective_type == 'overlap':
bins = len(bin_centers)
quad_profiles = []
for i in range(4):
bc, bv = self.cspad.intensity_profile(raw_image, n_bins=bins, quad=i)
quad_profiles.append(bv)
m = np.min([ a.shape for a in quad_profiles ])
quad_profiles = np.array([ a[:m] for a in quad_profiles ])
obj = - np.sum(np.product(quad_profiles, axis=0))
# old objective function : peak height
elif self.objective_type == 'peak_height':
obj = - bin_values.max()
if self.width_weight != 0.0:
obj += self.width_weight * self._simple_width(bin_values, bar=self.horizontal_cut)
if self.peak_weight != 0.0:
obj += self.peak_weight
else:
raise ValueError('No implemented objective_type: %s' % objective_type)
# ----------------------------------------------------------------------
print "objective value: %.4e" % obj
return obj
