# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""Regions helper functions.
Throughout Gammapy, we use `regions` to represent and work with regions.
https://astropy-regions.readthedocs.io
We might add in other conveniences and features here, e.g. sky coord contains
without a WCS (see "sky and pixel regions" in PIG 10), or some HEALPix integration.
TODO: before Gammapy v1.0, discuss what to do about ``gammapy.utils.regions``.
Options: keep as-is, hide from the docs, or to remove it completely
(if the functionality is available in ``astropy-regions`` directly.
"""
import operator
import numpy as np
from scipy.optimize import Bounds, minimize
from astropy import units as u
from astropy.coordinates import SkyCoord
from regions import (
CircleAnnulusSkyRegion,
CircleSkyRegion,
CompoundSkyRegion,
RectangleSkyRegion,
Regions,
)
__all__ = [
"compound_region_to_regions",
"make_concentric_annulus_sky_regions",
"make_orthogonal_rectangle_sky_regions",
"regions_to_compound_region",
]
def compound_region_center(compound_region):
"""Compute center for a CompoundRegion
The center of the compound region is defined here as the geometric median
of the individual centers of the regions. The geometric median is defined
as the point the minimises the distance to all other points.
Parameters
----------
compound_region : `CompoundRegion`
Compound region
Returns
-------
center : `~astropy.coordinates.SkyCoord`
Geometric median of the positions of the individual regions
"""
regions = compound_region_to_regions(compound_region)
if len(regions) == 1:
return regions[0].center
positions = SkyCoord([region.center for region in regions])
def f(x, coords):
"""Function to minimize"""
lon, lat = x
center = SkyCoord(lon * u.deg, lat * u.deg)
return np.sum(center.separation(coords).deg)
ra, dec = positions.icrs.ra.wrap_at("180d").deg, positions.icrs.dec.deg
ub = np.array([np.max(ra), np.max(dec)])
lb = np.array([np.min(ra), np.min(dec)])
if np.all(ub == lb):
bounds = None
else:
bounds = Bounds(ub=ub, lb=lb)
result = minimize(
f,
x0=[np.mean(ra), np.mean(dec)],
args=(positions,),
bounds=bounds,
method="L-BFGS-B",
)
return SkyCoord(result.x[0], result.x[1], frame="icrs", unit="deg")
[docs]def compound_region_to_regions(region):
"""Create list of regions from compound regions.
Parameters
----------
region : `~regions.CompoundSkyRegion` or `~regions.SkyRegion`
Compound sky region
Returns
-------
regions : `~regions.Regions`
List of regions.
"""
regions = Regions([])
if isinstance(region, CompoundSkyRegion):
if region.operator is operator.or_:
regions_1 = compound_region_to_regions(region.region1)
regions.extend(regions_1)
regions_2 = compound_region_to_regions(region.region2)
regions.extend(regions_2)
else:
raise ValueError("Only union operator supported")
else:
return Regions([region])
return regions
[docs]def regions_to_compound_region(regions):
"""Create compound region from list of regions, by creating the union.
Parameters
----------
regions : `~regions.Regions`
List of regions.
Returns
-------
compound : `~regions.CompoundSkyRegion` or `~regions.CompoundPixelRegion`
Compound sky region
"""
region_union = regions[0]
for region in regions[1:]:
region_union = region_union.union(region)
return region_union
class SphericalCircleSkyRegion(CircleSkyRegion):
"""Spherical circle sky region.
TODO: is this separate class a good idea?
- If yes, we could move it to the ``regions`` package?
- If no, we should implement some other solution.
Probably the alternative is to add extra methods to
the ``CircleSkyRegion`` class and have that support
both planar approximation and spherical case?
Or we go with the approach to always make a
TAN WCS and not have true cone select at all?
"""
def contains(self, skycoord, wcs=None):
"""Defined by spherical distance."""
separation = self.center.separation(skycoord)
return separation < self.radius
[docs]def make_orthogonal_rectangle_sky_regions(start_pos, end_pos, wcs, height, nbin=1):
"""Utility returning an array of regions to make orthogonal projections
Parameters
----------
start_pos : `~astropy.regions.SkyCoord`
First sky coordinate defining the line to which the orthogonal boxes made
end_pos : `~astropy.regions.SkyCoord`
Second sky coordinate defining the line to which the orthogonal boxes made
height : `~astropy.quantity.Quantity`
Height of the rectangle region.
wcs : `~astropy.wcs.WCS`
WCS projection object
nbin : int
Number of boxes along the line
Returns
-------
regions : list of `~regions.RectangleSkyRegion`
Regions in which the profiles are made
"""
pix_start = start_pos.to_pixel(wcs)
pix_stop = end_pos.to_pixel(wcs)
points = np.linspace(start=pix_start, stop=pix_stop, num=nbin + 1).T
centers = 0.5 * (points[:, :-1] + points[:, 1:])
coords = SkyCoord.from_pixel(centers[0], centers[1], wcs)
width = start_pos.separation(end_pos).to("rad") / nbin
angle = end_pos.position_angle(start_pos) - 90 * u.deg
regions = []
for center in coords:
reg = RectangleSkyRegion(
center=center, width=width, height=u.Quantity(height), angle=angle
)
regions.append(reg)
return regions
[docs]def make_concentric_annulus_sky_regions(
center, radius_max, radius_min=1e-5 * u.deg, nbin=11
):
"""Make a list of concentric annulus regions.
Parameters
----------
center : `~astropy.coordinates.SkyCoord`
Center coordinate
radius_max : `~astropy.units.Quantity`
Maximum radius.
radius_min : `~astropy.units.Quantity`
Minimum radius.
nbin : int
Number of boxes along the line
Returns
-------
regions : list of `~regions.RectangleSkyRegion`
Regions in which the profiles are made
"""
regions = []
edges = np.linspace(radius_min, u.Quantity(radius_max), nbin)
for r_in, r_out in zip(edges[:-1], edges[1:]):
region = CircleAnnulusSkyRegion(
center=center,
inner_radius=r_in,
outer_radius=r_out,
)
regions.append(region)
return regions