def synthetic_spectrum():
param = get_para()
x = np.arange(2000)
pileup_peak = ['Si_Ka1-Si_Ka1', 'Si_Ka1-Ce_La1']
elemental_lines = ['Ar_K', 'Fe_K', 'Ce_L', 'Pt_M'] + pileup_peak
elist, matv, area_v = construct_linear_model(x, param, elemental_lines, default_area=1e5)
return np.sum(matv, 1) #In case that y0 might be zero at certain points.
def calculate_profile(x, y, param, elemental_lines,
default_area=1e5):
"""
Calculate the spectrum profile based on given paramters. Use function
construct_linear_model from xrf_model.
Parameters
----------
x : array
channel array
y : array
spectrum intensity
param : dict
paramters
elemental_lines : list
such as Si_K, Pt_M
required_length : optional, int
the length of the array might change due to trim process, so
predifine the length to a given value.
default_area : float
default value for the gaussian area of each element
Returns
-------
x : array
trimmed energy range
temp_d : dict
dict of array
area_dict : dict
dict of area for elements and other peaks
"""
# Need to use deepcopy here to avoid unexpected change on parameter dict
fitting_parameters = copy.deepcopy(param)
total_list, matv, area_dict = construct_linear_model(x,
fitting_parameters,
elemental_lines,
default_area=default_area)
temp_d = {k: v for (k, v) in zip(total_list, matv.transpose())}
# add background
bg = snip_method(y,
fitting_parameters['e_offset']['value'],
fitting_parameters['e_linear']['value'],
fitting_parameters['e_quadratic']['value'],
width=fitting_parameters['non_fitting_values']['background_width'])
temp_d['background'] = bg
x_energy = (fitting_parameters['e_offset']['value']
+ fitting_parameters['e_linear']['value'] * x
+ fitting_parameters['e_quadratic']['value'] * x**2)
return x_energy, temp_d, area_dict
def synthetic_spectrum():
param = get_para()
x = np.arange(2000)
pileup_peak = ['Si_Ka1-Si_Ka1', 'Si_Ka1-Ce_La1']
user_peak = ['user_peak1']
elemental_lines = (['Ar_K', 'Fe_K', 'Ce_L', 'Pt_M'] + pileup_peak +
user_peak)
elist, matv, area_v = construct_linear_model(x, param, elemental_lines,
default_area=1e5)
# In case that y0 might be zero at certain points.
return np.sum(matv, 1)
def calculate_profile(x, y, param, elemental_lines,
default_area=1e5):
"""
Calculate the spectrum profile based on given paramters. Use function
construct_linear_model from xrf_model.
Parameters
----------
x : array
channel array
y : array
spectrum intensity
param : dict
paramters
elemental_lines : list
such as Si_K, Pt_M
required_length : optional, int
the length of the array might change due to trim process, so
predifine the length to a given value.
default_area : float
default value for the gaussian area of each element
Returns
-------
x : array
trimmed energy range
temp_d : dict
dict of array
area_dict : dict
dict of area for elements and other peaks
"""
# Need to use deepcopy here to avoid unexpected change on parameter dict
fitting_parameters = copy.deepcopy(param)
total_list, matv, area_dict = construct_linear_model(x,
fitting_parameters,
elemental_lines,
default_area=default_area)
temp_d = {k: v for (k, v) in zip(total_list, matv.transpose())}
# add background
bg = snip_method(y,
fitting_parameters['e_offset']['value'],
fitting_parameters['e_linear']['value'],
fitting_parameters['e_quadratic']['value'],
width=fitting_parameters['non_fitting_values']['background_width'])
temp_d['background'] = bg
x_energy = (fitting_parameters['e_offset']['value']
+ fitting_parameters['e_linear']['value'] * x
+ fitting_parameters['e_quadratic']['value'] * x**2)
return x_energy, temp_d, area_dict
def test_pixel_fit_multiprocess():
param = get_para()
y0 = synthetic_spectrum()
x = np.arange(len(y0))
pileup_peak = ['Si_Ka1-Si_Ka1', 'Si_Ka1-Ce_La1']
user_peak = ['user_peak1']
elemental_lines = (['Ar_K', 'Fe_K', 'Ce_L', 'Pt_M'] + pileup_peak +
user_peak)
default_area = 1e5
elist, matv, area_v = construct_linear_model(x,
param,
elemental_lines,
default_area=default_area)
exp_data = np.zeros([2, 1, len(y0)])
for i in range(exp_data.shape[0]):
exp_data[i, 0, :] = y0
results = fit_pixel_multiprocess_nnls(exp_data, matv, param, use_snip=True)
# output area of dict
result_map = calculate_area(elist,
matv,
results,
param,
first_peak_area=True)
# compare input list and output elemental list
assert_array_equal(elist, elemental_lines + ['compton', 'elastic'])
# Total len includes all the elemental list, compton, elastic and
# two more items, which are summed area of background and r-squared
total_len = len(elist) + 2
assert_array_equal(results.shape, [2, 1, total_len])
# same exp data should output same results
assert_array_equal(results[0, :, :], results[1, :, :])
for k, v in six.iteritems(result_map):
assert_equal(v[0, 0], v[1, 0])
if k in ['snip_bkg', 'r_squared']:
# bkg is not a fitting parameter, and r_squared is just a
# statistical output.
# Only compare the fitting parameters, such as area of each peak.
continue
# compare with default value 1e5, and get difference < 1%
assert abs(v[0, 0] * 0.01 - default_area) / default_area < 1e-2
def test_pixel_fit_multiprocess():
param = get_para()
y0 = synthetic_spectrum()
x = np.arange(len(y0))
pileup_peak = ['Si_Ka1-Si_Ka1', 'Si_Ka1-Ce_La1']
user_peak = ['user_peak1']
elemental_lines = (['Ar_K', 'Fe_K', 'Ce_L', 'Pt_M'] + pileup_peak +
user_peak)
default_area = 1e5
elist, matv, area_v = construct_linear_model(x, param, elemental_lines,
default_area=default_area)
exp_data = np.zeros([2, 1, len(y0)])
for i in range(exp_data.shape[0]):
exp_data[i, 0, :] = y0
results = fit_pixel_multiprocess_nnls(exp_data, matv, param,
use_snip=True)
# output area of dict
result_map = calculate_area(elist, matv, results,
param, first_peak_area=True)
# compare input list and output elemental list
assert_array_equal(elist, elemental_lines+['compton', 'elastic'])
# Total len includes all the elemental list, compton, elastic and
# two more items, which are summed area of background and r-squared
total_len = len(elist) + 2
assert_array_equal(results.shape, [2, 1, total_len])
# same exp data should output same results
assert_array_equal(results[0, :, :], results[1, :, :])
for k, v in six.iteritems(result_map):
assert_equal(v[0, 0], v[1, 0])
if k in ['snip_bkg', 'r_squared']:
# bkg is not a fitting parameter, and r_squared is just a
# statistical output.
# Only compare the fitting parameters, such as area of each peak.
continue
# compare with default value 1e5, and get difference < 1%
assert_true(abs(v[0, 0] * 0.01 - default_area) / default_area < 1e-2)
