def test_refine_traces_curved_trace():
nx, ny = 1001, 1003
expected_trace = np.zeros((ny, nx), dtype=int)
trace_half_height = 5
x0 = 498
input_y_center = 2e-4 * (np.arange(nx) - x0)**2.0 + 0.01 * (np.arange(nx) -
x0) + 502.0
trace = np.round(input_y_center).astype(int)
for i in np.arange(nx, dtype=int):
for j in range(-trace_half_height, trace_half_height + 1):
expected_trace[trace[i] + j, i] = 1
x2d, y2d = np.meshgrid(np.arange(nx, dtype=int), np.arange(ny, dtype=int))
sigma = 1.5
y_center = np.array([input_y_center.copy() for i in range(ny)])
test_data = 1 / sigma / (2.0 * np.pi) * np.exp(
-0.5 * (y2d - y_center)**2.0 / sigma**2)
test_image = NRESObservationFrame([
CCDData(data=test_data,
uncertainty=np.zeros_like(test_data),
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
test_image.traces = np.ones_like(expected_trace)
refine_traces(test_image,
weights=test_image.data,
trace_half_height=trace_half_height)
np.testing.assert_equal(test_image.traces, expected_trace)
def two_order_image():
# generate 2 flat traces.
traces = np.zeros((60, 20))
traces[[10, 11, 12], :] = 1
# the second trace that does not span the image entirely
traces[[50, 51, 52], :] = 2
traces[[50, 51, 52], 0] = 0
traces[[50, 51, 52], -1] = 0
# generate test data with zero noise
data = np.ones_like(traces, dtype=float)
data[~np.isclose(traces, 0)] = 100.
uncertainty = 1. * data
wavelengths = (traces > 0).astype(float) * 5
image = NRESObservationFrame([
CCDData(data=data,
uncertainty=uncertainty,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
image.wavelengths = wavelengths
image.traces = traces
image.fibers = {'fiber': np.arange(2), 'order': np.arange(2)}
image.blaze = {
'id': np.arange(2),
'blaze': [np.arange(20), np.arange(20)],
'blaze_error': [np.arange(20), np.arange(20)]
}
return image
def test_refine_traces_offset_centroid():
nx, ny = 401, 403
x2d, y2d = np.meshgrid(np.arange(nx), np.arange(ny))
test_data, trace_centers, input_traces = make_simple_traces(nx, ny)
read_noise = 10.0
# Filter out the numerical noise
test_data[test_data < 1e-15] = 0.0
test_data += np.random.poisson(test_data)
test_data += np.random.normal(0.0, read_noise, size=test_data.shape)
uncertainty = np.sqrt(test_data + read_noise**2.0)
test_image = NRESObservationFrame([
CCDData(data=test_data,
uncertainty=uncertainty,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
test_image.traces = input_traces
input_context = context.Context({'TRACE_HALF_HEIGHT': 5})
stage = TraceRefiner(input_context)
output_image = stage.do_stage(test_image)
for trace_center in trace_centers:
# Make sure that the center +- 4 pixels is in the trace image
assert all(output_image.traces[np.abs(y2d - trace_center + 1) <= 4])
def test_background_subtraction_with_traces(seed):
nx, ny = 405, 403
x = np.arange(nx)
y = np.arange(ny)
X, Y = np.meshgrid(x, y)
noise_sigma = 1.0
input_background = 30 * np.exp(-(X - nx / 2.0)**2 / 300**2 -
(Y - ny / 2.0 - 50.0)**2 / 200**2)
test_data = input_background + np.random.normal(
0.0, noise_sigma, size=input_background.shape)
test_image = NRESObservationFrame([
CCDData(data=test_data,
uncertainty=np.ones((ny, nx)) * noise_sigma,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
test_image.traces = np.zeros((ny, nx))
for i in range(1, 8):
test_image.traces[40 * i:40 * i + 10] = i
input_context = context.Context({})
stage = BackgroundSubtractor(input_context)
output_image = stage.do_stage(test_image)
# Make sure our background estimation is good. This is roughly 4 counts which is not bad
# If we fully model the image, we can do better than this, but that becomes computationally prohibitive on a 4kx4k
# image
np.testing.assert_allclose(output_image.background,
input_background,
atol=5.0)
def test_profile_fit_with_noise_with_blaze():
nx, ny = 401, 403
input_profile, trace_centers, input_traces = make_simple_traces(nx,
ny,
blaze=True)
read_noise = 10.0
# Filter out the numerical noise
input_profile[input_profile < 1e-15] = 0.0
input_profile += np.random.poisson(input_profile)
input_profile += np.random.normal(0.0,
read_noise,
size=input_profile.shape)
uncertainty = np.sqrt(input_profile + read_noise**2.0)
image = NRESObservationFrame([
CCDData(data=input_profile.copy(),
uncertainty=uncertainty,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
image.traces = input_traces
input_context = context.Context({})
stage = ProfileFitter(input_context)
image = stage.do_stage(image)
for i in range(1, np.max(image.traces) + 1):
this_trace = get_trace_region(image.traces == i)
scale_factor = np.max(input_profile[this_trace]) / np.max(
image.profile[this_trace] * image.blaze[i - 1]['blaze'])
np.testing.assert_allclose(input_profile[this_trace],
image.profile[this_trace] *
image.blaze[i - 1]['blaze'] * scale_factor,
rtol=1.5e-2,
atol=1.0)
def test_do_stage_on_empty_features(self):
input_context = context.Context({})
image = NRESObservationFrame([CCDData(data=self.data, uncertainty=self.err,
meta={'OBJECTS': 'tung&tung&none'})], 'foo.fits')
image.traces = np.ones_like(self.data, dtype=int)
image.blaze = {'blaze': np.ones_like(self.data, dtype=int)}
stage = IdentifyFeatures(input_context)
stage.fwhm, stage.nsigma = self.sigma, 1.5
image = stage.do_stage(image)
assert len(image.features) == 0
def generate_image(self):
traces = np.array([[0, 0, 0], [1, 1, 1], [1, 1, 1], [0, 0, 0], [0, 0, 0], [2, 2, 2], [3, 3, 3]])
line_list = np.array([10, 11, 12])
pixel_positions = np.array([1, 2, 1, 2])
features = Table({'pixel': pixel_positions, 'id': np.array([1, 1, 2, 3]), 'order': np.array([1, 1, 2, 3]),
'wavelength': pixel_positions, 'centroid_err': np.ones_like(pixel_positions)})
image = NRESObservationFrame([CCDData(data=np.zeros((2, 2)), uncertainty=np.zeros((2, 2)),
meta={'OBJECTS': 'tung&tung&none'})], 'foo.fits')
image.features = features
image.traces = traces
image.line_list = line_list
return image
def test_do_stage_no_blaze(self):
input_context = context.Context({})
ccd_data = CCDData(data=self.data, uncertainty=self.err, meta={'OBJECTS': 'tung&tung&none'})
image = NRESObservationFrame([ccd_data], 'foo.fits')
image.traces = np.ones_like(self.data, dtype=int)
stage = IdentifyFeatures(input_context)
stage.fwhm, stage.nsigma = self.sigma, 0.5
image = stage.do_stage(image)
image.features.sort('pixel')
assert np.allclose(image.features['pixel'], self.xcoords, atol=0.001)
assert np.allclose(image.features['ycentroid'], self.ycoords, atol=0.001)
assert np.allclose(image.features['id'], 1)
def test_do_stage(self):
blaze_factor = 0.5
input_context = context.Context({})
ccd_data = CCDData(data=self.data, uncertainty=self.err, meta={'OBJECTS': 'tung&tung&none'})
image = NRESObservationFrame([ccd_data], 'foo.fits')
image.traces = np.ones_like(self.data, dtype=int)
image.blaze = {'blaze': blaze_factor * np.ones((1, self.data.shape[1]), dtype=int)}
stage = IdentifyFeatures(input_context)
stage.fwhm, stage.nsigma = self.sigma, 0.5
image = stage.do_stage(image)
image.features.sort('pixel')
assert np.allclose(image.features['corrected_flux'], image.features['flux'] / blaze_factor, rtol=1E-4)
assert np.allclose(image.features['pixel'], self.xcoords, atol=0.001)
assert np.allclose(image.features['ycentroid'], self.ycoords, atol=0.001)
assert np.allclose(image.features['id'], 1)
def five_hundred_square_image(maxflux,
number_traces,
trace_width,
read_noise=10,
seed=None):
traces = np.zeros((500, 500))
data = np.ones_like(traces, dtype=float)
profile = np.zeros_like(traces, dtype=float)
ix = np.arange(trace_width)
for i in range(1, number_traces + 1):
traces[40 * i:40 * i + trace_width, :] = i
for j in range(0, trace_width):
data[40 * i + j, :] += maxflux * np.exp(
(-1.) * (ix[j] - trace_width / 2.)**2 / (trace_width / 6.)**2)
for j in range(0, trace_width):
profile[40 * i + j, :] = data[40 * i + j, :] / np.sum(
data[40 * i:40 * i + trace_width, 0])
np.random.seed(seed=seed)
data += np.random.poisson(data)
data += np.random.normal(0.0, read_noise, size=data.shape)
uncertainty = np.sqrt(data + read_noise**2)
wavelengths = np.ones_like(
traces
) * 5 # dummy wavelengths image that has values distinct from flux and traces.
image = NRESObservationFrame([
CCDData(data=data,
uncertainty=uncertainty,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
image.traces = traces
image.profile = profile
image.wavelengths = wavelengths
image.blaze = {
'id': np.arange(number_traces) + 1,
'blaze': [np.ones(traces.shape[1]) for i in range(number_traces)],
'blaze_error':
[np.ones(traces.shape[1]) for i in range(number_traces)]
}
image.fibers = {
'fiber': np.arange(number_traces) + 1,
'order': np.arange(number_traces) + 1
}
return image
def test_refine_traces_with_previous_trace():
nx, ny = 401, 403
num_traces = 3
read_noise = 10.0
test_data = np.zeros((ny, nx))
x2d, y2d = np.meshgrid(np.arange(nx), np.arange(ny))
trace_half_width = 6
trace_centers = []
y_0s = [100, 200, 300]
blaze_function = 1 - 1e-5 * (x2d - nx / 2.)**2
input_traces = np.zeros((ny, nx), dtype=np.int)
for i in range(num_traces):
trace_centers.append(5e-4 * (np.arange(nx) - nx / 2.)**2 + y_0s[i])
test_data += gaussian(y2d, trace_centers[i], 2,
a=10000.0) * blaze_function
input_traces[np.abs(y2d -
trace_centers[i]) <= trace_half_width] = i + 1
# Filter out the numerical noise
test_data[test_data < 1e-15] = 0.0
test_data += np.random.poisson(test_data)
test_data += np.random.normal(0.0, read_noise, size=test_data.shape)
uncertainty = np.sqrt(test_data + read_noise**2.0)
test_image = NRESObservationFrame([
CCDData(data=test_data,
uncertainty=uncertainty,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
test_image.traces = input_traces
input_context = context.Context({'TRACE_HALF_HEIGHT': 5})
stage = TraceRefiner(input_context)
output_image = stage.do_stage(test_image)
for trace_center in trace_centers:
# Make sure that the center +- 4 pixels is in the trace image
assert all(output_image.traces[np.abs(y2d - trace_center) <= 4])
assert np.isclose(num_traces, output_image.num_traces)
def test_profile_fit_without_noise_without_blaze():
nx, ny = 401, 403
input_profile, trace_centers, input_traces = make_simple_traces(
nx, ny, blaze=False)
# Filter out the numerical noise
input_profile[input_profile < 1e-15] = 0.0
uncertainty = np.sqrt(input_profile)
image = NRESObservationFrame([
CCDData(data=input_profile.copy(),
uncertainty=uncertainty,
meta={'OBJECTS': 'tung&tung&none'})
], 'foo.fits')
image.traces = input_traces
input_context = context.Context({})
stage = ProfileFitter(input_context)
image = stage.do_stage(image)
scale_factor = np.max(input_profile) / np.max(image.profile)
assert np.allclose(input_profile[input_traces != 0],
image.profile[input_traces != 0] * scale_factor,
rtol=5e-3,
atol=1.0)