示例#1
0
def analyse_source(source, cube, plot=False, return_fit_params=False):
    """
    Convenience method to spatially analyse a source.
    A 30x30 pixels 'flux map' is build from the sum of a few frames around each detected lines for the source.

    Two analysis are then performed:

    * Aperture photometry from the center of the source, to estimate a flux growth function and fit it with a custom erf function.
    * A Gaussian 2D fit of the PSF on the flux map



    This method can be used in a parallel process.
    Parameters
    ----------
    source : :class:`~pandas:pandas.Series`
        A row from a :class:`~pandas:pandas.DataFrame` containing detected sources. Should have columns ``xpos``, ``ypos`` (Not astropy convention), ``velocity``, ``*_detected`` whare * is a line name, containing True or False for each line.
    cube : :class:`~ORCS:orcs.process.SpectralCube`
        SpectralCube instance where we are looking at the source
    plot : bool, Default = False
        (Optional) If True, the two fits are plotted
    return_fit_params : bool, Default = False
        (Optional) If True, returns the full fits parameters

    Returns
    -------
    res : dict
        A dictionnary containing all the relevant fitted quantities.

        +-----------------------+---------------------------------------------------------------------------------------------------+
        |Parameter              |Description                                                                                        |
        +=======================+===================================================================================================+
        |flux_map_ks_pvalue     |Estimates the 'randomness' of the flux map, i.e if it's just noise or if we actually have a signal |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |flux_r                 |Flux at different radius *r*                                                                       |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |flux_err_r             |Flux error varying with *r*                                                                        |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |erf_amplitude          |Amplitude estimated from erf fit                                                                   |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |erf_amplitude_err      |Amplitude error                                                                                    |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |erf_xfwhm              |x-axis fwhm from erf fit                                                                           |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |erf_yfwhm              |y-axis fwhm from erf fit                                                                           |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |erf_fwhm               |Fwhm defined as *r* at which half of the max flux is reached                                       |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |flux_fraction_3        |Ratio between flux measured at 3 pixels from the center and max flux                               |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |model_flux_fraction_15 |Ratio between estimated flux at 15 pixels from the center and max flux                             |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |modeled_flux_r         |Modeled flux varying with *r*                                                                      |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |psf_snr                |Ratio between amplitude of the 2D fit and noise in the flux map                                    |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |psf_amplitude          |Amplitude of the 2D fit                                                                            |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |psf_xfwhm              |x-axis fwhm from 2D fit                                                                            |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |psf_yfwhm              |y-axis fwhm from 2D fit                                                                            |
        +-----------------------+---------------------------------------------------------------------------------------------------+
        |psf_ks_pvalue          |Randomness of the residuals map                                                                    |
        +-----------------------+---------------------------------------------------------------------------------------------------+



    """
    result = {}
    try:
        from astropy.stats import sigma_clipped_stats, gaussian_sigma_to_fwhm
        from sitelle.constants import SN2_LINES, SN3_LINES
        from sitelle.region import centered_square_region
        from orb.utils.spectrum import line_shift
        from orb.core import Lines

        filter_name = cube.params.filter_name

        if filter_name == 'SN2':
            LINES = SN2_LINES
        elif filter_name == 'SN3':
            LINES = SN3_LINES
        else:
            raise ValueError(filter_name)

        ## We build a flux map of the detected lines
        try:
            detected_lines = [
                line_name for line_name in LINES
                if source['%s_detected' %
                          line_name.lower().replace('[', '').replace(']', '')]
            ]
        except KeyError as e:
            raise ValueError('No columns *_detected in the source')
        if detected_lines == []:
            return pd.Series(result)

        x, y = source.as_matrix(['xpos', 'ypos']).astype(int)
        big_box = centered_square_region(x, y, 30)
        medium_box = centered_square_region(15, 15, 5)
        small_box = centered_square_region(15, 15, 3)
        data = cube._extract_spectra_from_region(big_box, silent=True)
        mask = np.ones((30, 30))
        mask[medium_box] = 0
        bkg_spec = np.nanmedian(data[np.nonzero(mask)], axis=0)
        data -= bkg_spec

        axis = cube.params.base_axis
        spec = np.nansum(data[small_box], axis=0)

        line_pos = np.atleast_1d(
            Lines().get_line_cm1(detected_lines) +
            line_shift(source['velocity'],
                       Lines().get_line_cm1(detected_lines),
                       wavenumber=True))
        pos_min = line_pos - cube.params.line_fwhm
        pos_max = line_pos + cube.params.line_fwhm
        pos_index = np.array([[
            np.argmin(np.abs(axis - pos_min[i])),
            np.argmin(np.abs(axis - pos_max[i]))
        ] for i in range(pos_min.shape[0])])

        bandpass_size = 0
        flux_map = np.zeros(data.shape[:-1])
        for line_detection in pos_index:
            bandpass_size += line_detection[1] - line_detection[0]
            flux_map += np.nansum(data[:, :,
                                       line_detection[0]:line_detection[1]],
                                  axis=-1)

        _, _, std_map = sigma_clipped_stats(data, axis=-1)
        flux_noise_map = np.sqrt(bandpass_size) * std_map

        #Test for randomness of the flux_map
        from scipy import stats
        result['flux_map_ks_pvalue'] = stats.kstest(
            (flux_map / flux_noise_map).flatten(), 'norm').pvalue

        #Fit of the growth function
        from photutils import RectangularAperture
        from scipy.special import erf
        from scipy.optimize import curve_fit

        try:
            _x0 = source['xcentroid'] - x + 15.
            _y0 = source['ycentroid'] - y + 15.
        except:
            _x0 = source['xpos'] - x + 15.
            _y0 = source['ypos'] - y + 15.

        flux_r = [0.]
        flux_err_r = [np.nanmin(flux_noise_map)]

        r_max = 15
        r_range = np.arange(1, r_max + 1)
        for r in r_range:
            #             aper = CircularAperture((_x0,_y0), r)
            aper = RectangularAperture((_x0, _y0), r, r, 0)
            flux_r.append(aper.do_photometry(flux_map)[0][0])
            flux_err_r.append(
                np.sqrt(aper.do_photometry(flux_noise_map**2)[0][0]))

        flux_r = np.atleast_1d(flux_r)
        flux_err_r = np.atleast_1d(flux_err_r)

        result['flux_r'] = flux_r
        result['flux_err_r'] = flux_err_r
        try:

            def model(r, x0, y0, sx, sy, A):
                return A * erf((r / 2. - x0) / (2 * sx * np.sqrt(2))) * erf(
                    (r / 2. - y0) / (2 * sy * np.sqrt(2)))

            R = np.arange(r_max + 1)
            p, cov = curve_fit(model,
                               R,
                               flux_r,
                               p0=[0, 0, 1.5, 1.5,
                                   flux_map.max()],
                               bounds=([-2, -2, -np.inf, -np.inf, -np.inf],
                                       [2, 2, np.inf, np.inf, np.inf]),
                               sigma=flux_err_r,
                               absolute_sigma=True,
                               maxfev=10000)
            if (p[2] < 0) != (p[3] < 0):
                if p[-1] < 0:
                    p[-1] = -p[-1]
                    if p[2] < 0:
                        p[2] = -p[2]
                    elif p[3] < 0:
                        p[3] = -p[3]
            if plot:
                f, ax = plt.subplots()
                ax.plot(R, model(R, *p), label='Fit')
                ax.errorbar(R, flux_r, flux_err_r, label='Flux')
                ax.set_ylabel('Flux')
                ax.set_xlabel('Radius from source')
                ax.legend()

            from scipy.optimize import bisect
            fwhm = bisect(lambda x: model(x, *p) - p[-1] / 2, 0.1, 10)
            result['erf_amplitude'] = p[-1]
            result['erf_amplitude_err'] = np.sqrt(np.diag(cov))[-1]
            result['erf_xfwhm'] = gaussian_sigma_to_fwhm * p[2]
            result['erf_yfwhm'] = gaussian_sigma_to_fwhm * p[3]
            result['erf_ks_pvalue'] = stats.kstest(
                (flux_r - model(R, *p)) / flux_err_r, 'norm').pvalue
            result['erf_fwhm'] = fwhm

            result['flux_fraction_3'] = flux_r[3] / p[-1]
            result['model_flux_fraction_15'] = model(R, *
                                                     p)[r_range[-1]] / p[-1]

            result['modeled_flux_r'] = model(R, *p)

        except Exception as e:
            print(e)
            pass

        ## 2D fit of the PSF
        from astropy.modeling import models, fitting

        fitter = fitting.LevMarLSQFitter()
        X, Y = np.mgrid[:30, :30]

        flux_std = np.nanmean(flux_noise_map)

        gauss_model = models.Gaussian2D(amplitude=np.nanmax(flux_map /
                                                            flux_std),
                                        x_mean=_y0,
                                        y_mean=_x0)
        gauss_model.bounds['x_mean'] = (14, 16)
        gauss_model.bounds['y_mean'] = (14, 16)
        gauss_fit = fitter(gauss_model, X, Y, flux_map / flux_std)

        if plot is True:
            f, ax = plt.subplots(1, 3, figsize=(8, 3))
            v_min = np.nanmin(flux_map)
            v_max = np.nanmax(flux_map)
            plot_map(flux_map, ax=ax[0], cmap='RdBu_r', vmin=v_min, vmax=v_max)
            ax[0].set_title("Data")
            plot_map(gauss_fit(X, Y) * flux_std,
                     ax=ax[1],
                     cmap='RdBu_r',
                     vmin=v_min,
                     vmax=v_max)
            ax[1].set_title("Model")
            plot_map(flux_map - gauss_fit(X, Y) * flux_std,
                     ax=ax[2],
                     cmap='RdBu_r',
                     vmin=v_min,
                     vmax=v_max)
            ax[2].set_title("Residual")

        result['psf_snr'] = gauss_fit.amplitude[0]
        result['psf_amplitude'] = flux_std * gauss_fit.amplitude[
            0] * 2 * np.pi * gauss_fit.x_stddev * gauss_fit.y_stddev
        result['psf_xfwhm'] = gauss_fit.x_fwhm
        result['psf_yfwhm'] = gauss_fit.y_fwhm
        normalized_res = (flux_map -
                          gauss_fit(X, Y) * flux_std) / flux_noise_map
        result['psf_ks_pvalue'] = stats.kstest(normalized_res.flatten(),
                                               'norm').pvalue
        if return_fit_params:
            return pd.Series(result), p, gauss_fit
        else:
            return pd.Series(result)
    except Exception as e:
        print e
        return pd.Series(result)
示例#2
0
def do_Rect_phot(pos, FWHM, trail, ap_min=3., ap_factor=1.5, \
                 win=None, wout=None, hout=None,\
                 sky_nsigma=3., sky_iter=10 ):
    if win == None:
        win  = 4 * FWHM + trail
    if wout == None:
        wout = 8 * FWHM + trail
    if hout == None:
        hout = 8 * FWHM
    N = len(pos)
    if pos.ndim == 1:
        N = 1
    theta    = give_theta(number_of_stars=5)
    an       = RectAn(pos, w_in=win, w_out=wout, h_out=hout, theta=theta)
    ap_size  = np.max([ap_min, ap_factor*FWHM])
    aperture = RectAp(pos, w=(trail+ap_size), h=ap_size, theta=theta)
    flux     = aperture.do_photometry(image_reduc, method='exact')[0]
    # do phot and get sum from aperture. [0] is sum and [1] is error.
    #For test:
#FWHM = FWHM_moffat.copy()
#trail=trail_len.copy()
#win  = 4 * FWHM + trail
#wout = 8 * FWHM + trail
#hout = 8 * FWHM
#N=len(pos_star_fit)
#an       = RectAn(pos_star_fit, w_in=win, w_out=wout, h_out=hout, theta=(theta+np.pi/2))
#ap_size  = 1.5*FWHM_moffat
#aperture = RectAp(pos_star_fit, w=(trail+ap_size), h=ap_size, theta=(theta+np.pi/2))
#flux     = aperture.do_photometry(image_reduc, method='exact')[0]
#plt.figure(figsize=(12,12))
#plt.imshow(image_reduc, origin='lower', vmin=-10, vmax=1000)
#an.plot(color='white')
#aperture.plot(color='red')
    flux_ss  = np.zeros(N)
    error    = np.zeros(N)
    for i in range(0, N):
        mask_an    = (an.to_mask(method='center'))[i]
        #   cf: test = mask_an.cutout(image_reduc) <-- will make cutout image.
        sky_an     = mask_an.apply(image_reduc)
        all_sky    = sky_an[np.nonzero(sky_an)]
        # only annulus region will be saved as np.ndarray
        msky, stdev, nsky, nrej = sky_fit(all_sky, method='Mode', mode_option='sex')
        area       = aperture.area()
        flux_ss[i] = flux[i] - msky*area  # sky subtracted flux
        error[i]   = np.sqrt( flux_ss[i]/gain \
                           + area * stdev**2 \
                           + area**2 * stdev**2 / nsky )
        if inputs.star_img_save:
            from matplotlib import pyplot as plt
            mask_ap    = (aperture.to_mask(method='exact'))[i]
            star_ap_ss = mask_ap.apply(image_reduc-msky)
            sky_an_ss  = mask_an.apply(image_reduc-msky)
            plt.suptitle('{0}, Star ID={1} ({nsky:3d} {nrej:3d} {msky:7.2f} {stdev:7.2f})'.format(
                    inputs.filename, i, nsky=nsky, nrej=nrej, msky=msky, stdev=stdev ))
            ax1 = plt.subplot(1,2,1)
            im1 = ax1.imshow(sky_an_ss, origin='lower')
            plt.colorbar(im1, orientation='horizontal')
            ax2 = plt.subplot(1,2,2)
            im2 = ax2.imshow(star_ap_ss, origin='lower')
            plt.colorbar(im2, orientation='horizontal')
            plt.savefig('{0}.star{1}.png'.format(inputs.filename, i))
            plt.clf()
        if pos.ndim > 1:
            print('\t[{x:7.2f}, {y:7.2f}], {nsky:3d} {nrej:3d} {msky:7.2f} {stdev:7.2f} {flux:7.1f} {ferr:3.1f}'.format(\
                          x=pos[i][0], y=pos[i][1], \
                          nsky=nsky, nrej=nrej, msky=msky, stdev=stdev,\
                          flux=flux_ss[i], ferr=error[i]))
    return flux_ss, error