Camera Geometries#

The CameraGeometry provides an easy way to work with images or data cubes related to Cherenkov Cameras. In ctapipe, a camera image is simply a flat 1D array (or 2D if time information is included), where there is one value per pixel. Of course, to work with such an array, one needs spatial information about how the pixels are laid out. Since CTA has at least 6 different camera types, and may have multiple versions of each as revisions are made, it is necessary to have a common way to describe all cameras.

So far there are several ways to construct a CameraGeometry:

  • EventSource instances have a subarray attribute, e.g. to obtain the geometry for the telescope with id 1, use:[1].camera.geometry. The TableLoader instance also has the .subarray attribute.

  • use the CameraGeometry constructor, where one has to specify all necessary information (pixel positions, types, areas, etc)

  • load it from a pre-written file (which can be in any format supported by astropy.table, as long as that format allows for header-keywords as well as table entries.

Once loaded, the CameraGeometry object gives you access the pixel positions, areas, neighbors, and shapes.

CameraGeometry is used by most image processing algorithms in the ctapipe.image module, as well as displays in the ctapipe.visualization module.


You can write out a CameraGeometry by using the CameraGeometry.to_table() method to turn it into an astropy.table.Table, and then call its write function. Reading it back in can be done with from_table()

geom = ~ctapipe.instrument.CameraGeometry(...)  # constructed elsewhere

geom.to_table().write('mycam.fits.gz') # FITS output
geom.to_table().write('mycam.h5', path='/cameras/mycam') # hdf5 output
geom.to_table().write('mycam.ecsv', format='ascii.ecsv') # text table

# later read back in:

geom = ~ctapipe.instrument.CameraGeometry.from_table('mycam.ecsv', format='ascii.ecsv')
geom = ~ctapipe.instrument.CameraGeometry.from_table('mycam.fits.gz')
geom = ~ctapipe.instrument.CameraGeometry.from_table('mycam.h5', path='/cameras/mycam')

A Note On Pixel Neighbors#

The CameraGeometry object provides two pixel-neighbor representations: a neighbor adjacency list (in the neighbors attribute) and a pixel adjacency matrix (in the neighbor_matrix attribute). The former is a list of lists, where element i is a list of neighbors j of the i*th pixel. The latter is a 2D matrix where row *i is a boolean mask of pixels that are neighbors. It is not necessary to load or specify either of these neighbor representations when constructing a CameraGeometry, since they will be computed on-the-fly if left blank, using a KD-tree nearest-neighbor algorithm.

It is recommended that all algorithms that need to be computationally fast use the neighbor_matrix attribute, particularly in conjunction with numpy operations, since it is quite speed-efficient.


from matplotlib import pyplot as plt

from ctapipe.instrument import SubarrayDescription

subarray ="dataset://gamma_prod5.simtel.zst")
geom =[1].camera.geometry

plt.figure(figsize=(8, 3))
plt.subplot(1, 2, 1)
plt.imshow(geom.neighbor_matrix, origin="lower")
plt.title("Pixel Neighbor Matrix")

plt.subplot(1, 2, 2)
plt.scatter(geom.pix_x, geom.pix_y)
plt.title("Pixel Positions")

(Source code, png, hires.png, pdf)


See also ctapipe.image.tailcuts_clean() and ctapipe.image.dilate() for usage examples.

Reference/API# Module#

Utilities for reading or working with Camera geometry files


CameraGeometry(name, pix_id, pix_x, pix_y, ...)

CameraGeometry is a class that stores information about a Cherenkov Camera that us useful for imaging algorithms and displays.



Supported Pixel Shapes Enum