def test_build_fit_matrix_identity():
i = np.identity(2)
r = linearfit.build_fit_matrix(0) # also test that default scale value = 1
assert np.allclose(i, r, rtol=0, atol=_ATOL)
r = linearfit.build_fit_matrix((0, 0), (1, 1))
assert np.allclose(i, r, rtol=0, atol=_ATOL)
def test_align_wcs_refcat_from_imcat(mock_fits_wcs, enforce):
shift = (12, -34)
rot = (15, 17)
scale = (1.0123, 0.9876)
crpix = mock_fits_wcs.wcs.crpix - 1
xy = 1024 * np.random.random((100, 2))
m = build_fit_matrix(rot, scale)
xyr = np.dot(xy - crpix, m) + crpix + shift
imcat = Table(xy, names=('x', 'y'))
refcat = Table(xyr, names=('x', 'y'))
tpwcs1 = FITSWCS(mock_fits_wcs, meta={'catalog': imcat})
tpwcs2 = copy.deepcopy(tpwcs1)
reftpwcs = FITSWCS(mock_fits_wcs, meta={'catalog': refcat})
input_catalogs = [reftpwcs, tpwcs1, tpwcs2]
status = align_wcs(
input_catalogs, refcat=None, fitgeom='general',
match=None, enforce_user_order=enforce, expand_refcat=True
)
assert status
for cat in input_catalogs:
if cat.meta['fit_info']['status'] == 'REFERENCE':
if enforce:
assert cat is reftpwcs
continue
assert cat.meta['fit_info']['status'] == 'SUCCESS'
def test_align_wcs_simple_ref_image_general(shift, rot, scale, fitgeom,
weighted, mock_fits_wcs):
xy = 1024 * np.random.random((100, 2))
if weighted:
w = np.ones((100, 1))
xy = np.hstack((xy, w))
names = ('x', 'y', 'weight')
else:
names = ('x', 'y')
m = build_fit_matrix(rot, scale)
xyr = np.dot(xy[:, :2], m.T) + shift
imcat = Table(xy, names=names)
radec = mock_fits_wcs.wcs_pix2world(xyr, 0)
if weighted:
radec = np.hstack((radec, w))
names = ('RA', 'DEC', 'weight')
else:
names = ('RA', 'DEC')
refcat = Table(radec, names=names)
tpwcs = FITSWCS(mock_fits_wcs, meta={'catalog': imcat})
status = align_wcs(tpwcs, refcat, fitgeom=fitgeom, match=None)
assert status
assert tpwcs.meta['fit_info']['status'] == 'SUCCESS'
assert tpwcs.meta['fit_info']['fitgeom'] == fitgeom
assert np.allclose(tpwcs.meta['fit_info']['shift'], shift)
assert np.allclose(tpwcs.meta['fit_info']['matrix'], m)
assert np.allclose(tpwcs.meta['fit_info']['rot'], rot)
assert tpwcs.meta['fit_info']['proper']
assert np.allclose(tpwcs.meta['fit_info']['scale'], scale)
assert tpwcs.meta['fit_info']['rmse'] < 1.0e-8
def test_align_wcs_simple_twpwcs_ref(mock_fits_wcs):
shift = (12, -34)
rot = (15, 17)
scale = (1.0123, 0.9876)
xy = 1024 * np.random.random((100, 2))
m = build_fit_matrix(rot, scale)
xyr = np.dot(xy, m.T) + shift
imcat = Table(xy, names=('x', 'y'))
refcat = Table(xyr, names=('x', 'y'))
tpwcs = FITSWCS(mock_fits_wcs, meta={'catalog': imcat})
reftpwcs = FITSWCS(mock_fits_wcs, meta={'catalog': refcat})
status = align_wcs(tpwcs,
reftpwcs,
ref_tpwcs=tpwcs,
fitgeom='general',
match=None)
assert status
assert tpwcs.meta['fit_info']['status'] == 'SUCCESS'
assert tpwcs.meta['fit_info']['fitgeom'] == 'general'
assert np.allclose(tpwcs.meta['fit_info']['shift'], shift)
assert np.allclose(tpwcs.meta['fit_info']['matrix'], m)
assert np.allclose(tpwcs.meta['fit_info']['rot'], rot)
assert tpwcs.meta['fit_info']['proper']
assert np.allclose(tpwcs.meta['fit_info']['scale'], scale)
assert tpwcs.meta['fit_info']['rmse'] < 1.0e-8
def mock_jwst_wcs():
cd = build_fit_matrix((36, 47), 1e-5)
w = make_mock_jwst_wcs(v2ref=123.0,
v3ref=500.0,
roll=115.0,
crpix=[512.0, 512.0],
cd=cd,
crval=[82.0, 12.0])
return w
def ideal_large_data(request):
# rscale data with proper rotations and no noise
uv = np.random.random((_LARGE_SAMPLE_SIZE, 2))
# assume an image size of 4096x2048:
uv[:, 0] *= 2047.0
uv[:, 1] *= 4095.0
# rotation angle(s):
angle = 360.0 * np.random.random() # 0 ... 360
if request.param == 'shifts':
angle = (0, 0)
scale = (1, 1)
proper = True
transform = 'shift'
elif request.param == 'rscale':
angle = (angle, angle)
scale = 2 * (0.8 + 0.4 * np.random.random(), ) # 0.8 ... 1.2
proper = True
transform = 'rscale'
elif request.param == 'rscale-flip-x':
angle = ((angle + 180.0) % 360.0, angle)
scale = 2 * (0.8 + 0.4 * np.random.random(), ) # 0.8 ... 1.2
proper = False
transform = 'rscale'
elif request.param == 'rscale-flip-y':
angle = (angle, (angle + 180.0) % 360.0)
scale = 2 * (0.8 + 0.4 * np.random.random(), ) # 0.8 ... 1.2
proper = False
transform = 'rscale'
elif request.param == 'affine':
# rotation angles:
offset = 150.0 * (np.random.random() - 0.5) # -75 ... 75
offset += 180.0 * np.random.choice([0.0, 1.0]) # add random axis flip
angle = (angle, (angle + offset) % 360.0)
# scales:
scale = 0.8 + 0.4 * np.random.random(2) # 0.8 ... 1.2
# proper:
rad = np.deg2rad(angle)
proper = (np.prod(np.cos(rad)) + np.prod(np.sin(rad))) > 0
transform = 'general'
shift = 200.0 * (np.random.random(2) - 0.5) # -100 ... +100
rmat = linearfit.build_fit_matrix(angle, scale)
skew = angle[1] - angle[0]
# apply rscale
xy = np.dot(uv, rmat.T) + shift
return uv, xy, angle, scale, shift, rmat, proper, skew, transform
def mock_fits_wcs():
cd = build_fit_matrix((36, 47), 1e-5)
w = fitswcs.WCS(naxis=2)
w.wcs.cd = cd
w.wcs.crval = [82.0, 12.0]
w.wcs.crpix = [512.0, 512.0]
w.wcs.ctype = ['RA---TAN', 'DEC--TAN']
w.pixel_shape = [1024, 2048]
w.pixel_bounds = ((-0.5, 1024 - 0.5), (-0.5, 2048 - 0.5))
w.wcs.set()
return w
def test_build_fit_matrix_generalized(rot, scale):
i = np.identity(2)
m = linearfit.build_fit_matrix(rot, scale)
# check scale:
assert np.allclose(np.sqrt(np.sum(m**2, axis=0)),
scale,
rtol=0,
atol=_ATOL)
ms = np.diag(scale)
# check rotations:
mr = linearfit.build_fit_matrix(rot, 1)
mrinv = linearfit.build_fit_matrix(rot[::-1], 1).T
assert np.allclose(np.linalg.det(mr) * i,
np.dot(mr, mrinv),
rtol=0,
atol=_ATOL)
assert np.allclose(m, np.dot(mr, ms), rtol=0, atol=_ATOL)
def test_fitswcs_coord_transforms(mock_fits_wcs):
w = fitswcs.WCS(naxis=2)
w.wcs.cd = build_fit_matrix(0, 1e-5)
w.wcs.crval = [12, 24]
w.wcs.crpix = [500, 512]
w.wcs.ctype = ['RA---TAN', 'DEC--TAN']
w.pixel_shape = [1024, 2048]
w.wcs.set()
wc = tpwcs.FITSWCS(w)
wc.set_correction()
assert np.allclose(wc.det_to_world(499, 511), (12, 24), atol=_ATOL)
assert np.allclose(wc.world_to_det(12, 24), (499, 511), atol=_ATOL)
assert np.allclose(wc.det_to_tanp(499, 511), (0, 0), atol=_ATOL)
assert np.allclose(wc.tanp_to_det(0, 0), (499, 511), atol=_ATOL)
assert np.allclose(wc.world_to_tanp(12, 24), (0, 0), atol=1e-8)
assert np.allclose(wc.tanp_to_world(0, 0), (12, 24), atol=_ATOL)
def test_inv_order2():
feps = 1000 * np.finfo(np.float64).eps
# Generate a 2D rotation/scale/skew matrix + some random noise.
# The reason for this complicated approach is to avoid creating
# singular matrices.
angle1 = 360 * np.random.random()
angle2 = angle1 + 50 * (np.random.random() - 0.5)
scale1 = 0.7 + 0.6 * np.random.random()
scale2 = 0.7 + 0.6 * np.random.random()
a = linearfit.build_fit_matrix((angle1, angle2), (scale1, scale2))
# invert using numpy.linalg:
use_numpy = linalg._USE_NUMPY_LINALG_INV
linalg._USE_NUMPY_LINALG_INV = True
try:
x = linalg.inv(a)
r = np.identity(2) - np.dot(a, x)
# Use Morris Newman's formula to asses the quality of the inversion
# (see https://nvlpubs.nist.gov/nistpubs/jres/78B/jresv78Bn2p65_A1b.pdf
# January 3, 1974).
err = 2.0 * np.abs(np.dot(x, r)).max() / (1.0 - np.abs(r).max())
assert err < feps
finally:
linalg._USE_NUMPY_LINALG_INV = use_numpy
# invert using tweakwcs.linalg:
linalg._USE_NUMPY_LINALG_INV = False
try:
x = linalg.inv(a)
r = np.identity(2) - np.dot(a, x)
# Use Morris Newman's formula to asses the quality of the inversion
# (see https://nvlpubs.nist.gov/nistpubs/jres/78B/jresv78Bn2p65_A1b.pdf
# January 3, 1974).
err = 2.0 * np.abs(np.dot(x, r)).max() / (1.0 - np.abs(r).max())
assert err < feps
finally:
linalg._USE_NUMPY_LINALG_INV = use_numpy
def test_align_wcs_minobj(mock_fits_wcs):
shift = (12, -34)
rot = (15, 17)
scale = (1.0123, 0.9876)
crpix = mock_fits_wcs.wcs.crpix - 1
xy = 1024 * np.random.random((100, 2))
m = build_fit_matrix(rot, scale)
xyr = np.dot(xy - crpix, m) + crpix + shift
imcat = Table(xy, names=('x', 'y'))
refcat = Table(xyr, names=('x', 'y'))
tpwcs1 = FITSWCS(mock_fits_wcs, meta={'catalog': imcat})
reftpwcs = FITSWCS(mock_fits_wcs, meta={'catalog': refcat})
status = align_wcs([reftpwcs, tpwcs1],
refcat=None,
fitgeom='general',
minobj=200,
match=None,
enforce_user_order=True,
expand_refcat=False)
assert status
assert reftpwcs.meta['fit_info']['status'] == 'REFERENCE'
assert tpwcs1.meta['fit_info']['status'] == 'FAILED: not enough matches'
def test_build_fit_matrix_rot(rot):
i = np.identity(2)
m = linearfit.build_fit_matrix(rot)
minv = linearfit.build_fit_matrix(360 - rot)
assert np.allclose(i, np.dot(m, minv), rtol=0, atol=_ATOL)
