def flux_weighted_centroid(img, box_edge, init=None, to_plot=False):
"""
Compute the flux-weighted centroid as described on p. 105 of
Howell (2006).
inputs
------
img: array-like
box_edge: odd integer
number of pixels in a box edge
init: array-like, n=2, optional
initial position at which to center the fitting box.
if None, the central pixel of the image will be chosen.
to_plot: boolean
If True, produce a diagnostic plot of flux vs. pixel
Default = False
"""
if (box_edge % 2)==0:
logging.warning("even box edge given; increasing by one")
L = box_edge / 2
box_edge += 1
else:
L = (box_edge - 1) / 2
if init is None:
init = np.asarray(np.shape(img) / 2.0, int)
raw_init = init
else:
raw_init = np.copy(init)
init = np.asarray(np.round(init), int)
# print init, raw_init
sub_img = img[init[1]-L:init[1]+L+1, init[0]-L:init[0]+L+1]
Isum = np.sum(sub_img, axis=0)
Jsum = np.sum(sub_img, axis=1)
# print np.shape(img)
xedge = np.arange(np.shape(img)[1])[init[0]-L:init[0]+L+1]
yedge = np.arange(np.shape(img)[0])[init[1]-L:init[1]+L+1]
# print xedge
# print yedge
Ibar = (1.0 / box_edge) * np.sum(Isum)
Jbar = (1.0 / box_edge) * np.sum(Jsum)
# print Ibar, Jbar
Ibar_diff = Isum - Ibar
# print Ibar_diff, np.sum(Ibar_diff)
Jbar_diff = Jsum - Jbar
# print Jbar_diff, np.sum(Jbar_diff)
xc_top = np.sum((Ibar_diff * xedge)[Ibar_diff>0])
xc_bot = np.sum(Ibar_diff[Ibar_diff>0])
yc_top = np.sum((Jbar_diff * yedge)[Jbar_diff>0])
yc_bot = np.sum(Jbar_diff[Jbar_diff>0])
xc = xc_top / xc_bot
yc = yc_top / yc_bot
# logging.debug("init %.2f %.2f c %.2f %.2f", init[0], init[1], xc, yc)
if to_plot:
plt.figure(figsize=(8,10))
grid = (4,2)
fake_mask = np.ones_like(img)
ax1 = plt.subplot2grid(grid, (0,0), rowspan=2)
plot.stamp(img, fake_mask, ax=ax1)
ax2 = plt.subplot2grid(grid, (0,1), rowspan=2)
plot.stamp(img[init[0]-L:init[0]+L+1, init[1]-L:init[1]+L+1],
fake_mask[init[0]-L:init[0]+L+1, init[1]-L:init[1]+L+1],
ax=ax2)
ax3 = plt.subplot2grid(grid, (2,0), colspan=2)
ax3.step(xedge, Isum, lw=2, where="mid")
ax4 = plt.subplot2grid(grid, (3,0), colspan=2)
ax4.step(yedge, Jsum, lw=2, where="mid")
ax3.axhline(Ibar,color="g", ls=":", lw=3)
ax4.axhline(Jbar,color="g", ls=":", lw=3)
ax3.axvline(raw_init[0],color="k", ls="--", lw=2)
ax4.axvline(raw_init[1],color="k", ls="--", lw=2)
ax3.axvline(xc,color="r", lw=2)
ax4.axvline(yc,color="r", lw=2)
ax3.set_xlabel("X Pixel")
ax4.set_xlabel("Y Pixel")
ax3.set_ylabel("Flux")
ax3.set_ylabel("Flux")
plot.centroids(ax1, init=init, coords=(xc,yc))
ax2.set_xticklabels(np.append(xedge,xedge[-1]+1))
ax2.set_yticklabels(np.append(yedge,yedge[-1]+1))
# plt.suptitle("TEST",fontsize="large")
plt.tight_layout()
return np.array([xc, yc])
def run_one(filename, output_f=None, extract_companions=False, fw_box=None):
"""
Extract a light curve for one star.
Inputs
------
filename: string
filename for target pixel file
output_f: string (optional)
if provided, statistics from the extraction will be printed to this file
extract_companions: boolean (optional; default=False)
if true, extract light curves for any other DAOFIND objects
detected in the pixel stamp.
fw_box: odd integer (optional)
size of the box within which to calculate the flux-weighted centroid.
Must be odd. Defaults to the shortest dimension of the pixel stamp,
or 9 if larger than 9.
"""
# Clip part of the filename for use in saving outputs
outfilename = filename.split("/")[-1][:-14]
logging.info(outfilename)
# Retrieve the data
table, times, pixels, maskmap, maskheader, kpmag = tpf_io.get_data(filename)
# Use the RA/Dec from the header as the initial position for calculation
init = centroid.init_pos(maskheader)
logging.info("init %f %f", init[0], init[1])
# If not provided, calculate the maximum possible centroid box size.
# Maximum possible box is 9, unless set larger by the user with fw_box
if fw_box is None:
min_ax = np.argmin(np.shape(maskmap))
min_box = np.shape(maskmap)[min_ax]
if min_box>=9:
min_box = 9
logging.debug("min_box set to 9")
elif min_ax==0:
for row in maskmap:
row_len = len(np.where(row>0)[0])
if row_len < min_box:
min_box = row_len
logging.debug("new min_box %d", min_box)
else:
for i in range(np.shape(maskmap)[1]):
col = maskmap[:, i]
col_len = len(np.where(col>0)[0])
if col_len < min_box:
min_box = row_len
logging.debug("new min_box %d", min_box)
fw_box = (min_box / 2) * 2 + 1
logging.info("fw box %d",fw_box)
# Co-add all the sub-images and calculate the flux-weighted centroid
coadd = np.sum(pixels,axis=0)
coords = centroid.flux_weighted_centroid(coadd, fw_box, init=init)
# Background of the co-added stampl is just the sigma-clipped median
mean, median, std = sigma_clipped_stats(coadd, mask=(maskmap==0),
sigma=3.0, iters=3)
coadd_bkgd = median
# Warn the user if the calculated initial coordinates are too far
# from the initial coordinates (indicting that there's a brighter nearby
# star pulling the centroid off of the target)
if np.sqrt((coords[0] - init[0])**2 + (coords[1] - init[1])**2)>1:
logging.warning("Centroid (%f %f) far from init (%f %f)",
coords[0], coords[1], init[0], init[1])
else:
logging.debug("coords %f %f", coords[0], coords[1])
# Set keyword arguments for DAOFIND - trying to find anything other
# real sources in the image
dkwargs = {"fwhm":2.5, "threshold":coadd_bkgd,
"sharplo":0.01,"sharphi":5.0}
sources, n_sources = centroid.daofind_centroid(coadd, None, dkwargs)
if n_sources==0:
logging.info("bkgd: %f max: %f", coadd_bkgd, max(coadd.flatten()))
logging.debug("sources")
logging.debug(sources)
# Set the minimum and maximum aperture sizes for the extraction
# and the step in aperture sizes
ap_min, ap_max, ap_step = 2, 7, 0.5
radii = np.arange(ap_min, ap_max, ap_step)
# Always use circular apertures, I think ap_type is now defunct actually...
ap_type = "circ"
phot.make_circ_lc(pixels, maskmap, times, init, radii,
"lcs/{}.csv".format(outfilename), fw_box)
# get the EPIC ID
epic = outfilename.split("-")[0][4:]
logging.info(epic)
# Plot an informational plot with the co-added stamp, possibly a DSS image,
# centroid motion, and location on the K2 FOV
# plot.plot_four(epic, filename, coadd, maskmap, maskheader, init, coords,
# sources, ap=None, campaign=4)
# If an output filename is provided, save the output
if output_f is not None:
output_f.write("\n{},{}".format(outfilename,epic))
output_f.write(",{0:.6f},{1:.6f}".format(maskheader["RA_OBJ"],
maskheader["DEC_OBJ"]))
print type(init[0]), type(init[1])
print init[0].dtype, init[1].dtype
print init[0], init[1]
output_f.write(",{0:.6f},{1:.6f}".format(np.float64(init[0]),
np.float64(init[1])))
output_f.write(",{0:.2f},{1}".format(kpmag, n_sources))
output_f.write(",{0:.2f},{1:.6f}".format(dkwargs["fwhm"],
dkwargs["threshold"]))
output_f.write(",{0:.2f},{1:.2f}".format(dkwargs["sharplo"],
dkwargs["sharphi"]))
output_f.write(",{0:.2f},{1}".format(fw_box,ap_type))
output_f.write(",{0:.2f},{1:.2f},{2:.2f}".format(ap_min, ap_max,
ap_step))
# If desired, also extract light curves for companions
if extract_companions:
for i, source in enumerate(sources):
# Calculate the separation between this source and the target
sep = np.sqrt((init[0] - source["xcentroid"])**2 +
(init[1] - source["ycentroid"])**2)
# If the source is close to the center, just skip (it's the target)
if sep<3:
continue
# If the source is further away, compute a lightcurve for it.
sletter = alphas[i]
init2 = np.array([source["xcentroid"], source["ycentroid"]])
coords2 = centroid.flux_weighted_centroid(coadd, 3, init=init2)
radii = np.arange(0.5, (sep / 2.0) + 0.1 ,0.5)
ax = plot.stamp(coadd, maskmap)
plot.centroids(ax, init2, coords2, sources)
plot.apertures(ax, init2, radii)
plt.savefig("plot_outputs/{}{}_stamp.png".format(outfilename,
sletter))
plt.title(outfilename+sletter)
phot.make_circ_lc(pixels, maskmap, times, init2[::-1], radii,
"lcs/{}{}.csv".format(outfilename, sletter))
#plot.lcs("lcs/{}{}.csv".format(outfilename, sletter), epic=epic)
plt.close("all")
