# Masking Solar System Objects in Images **URL:** \ **Category:** Astropy\ **Created:** [November 8, 2023, 8:21pm UTC](https://community.openastronomy.org/t/masking-solar-system-objects-in-images/792 "2023-11-08T20:21:31Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![evidlo](https://dub1.discourse-cdn.com/flex005/user_avatar/community.openastronomy.org/evidlo/32/445_2.png) [@evidlo](https://community.openastronomy.org/u/evidlo)\ **Post date:** [November 8, 2023, 8:21pm UTC](https://community.openastronomy.org/t/masking-solar-system-objects-in-images/792/1 "2023-11-08T20:21:31Z")
I’m new to Astropy and would appreciate some help. I’m trying to mask out pixels in images from a space telescope whose lines of sight pass through e.g. the Lunar or Earth discs. I’ve found [this sunpy example](https://docs.sunpy.org/en/stable/generated/gallery/computer_vision_techniques/mask_disk.html) which is quite close to what I’m looking for, but needs to work for the Earth and Moon. Any guidance would be appreciated. What I have: - image data in ndarray - instrument orientation unit vectors (3x3 matrix in GCRS) - instrument position (3x1 matrix in GCRS) and date - instrument FOV What I want: - boolean array mask whether each pixel’s LOS intersects a sphere of size `r` centered around a SkyCoord returned by `get_body()` ---
**Author:** ![ayshih](https://dub1.discourse-cdn.com/flex005/user_avatar/community.openastronomy.org/ayshih/32/84_2.png) [@ayshih](https://community.openastronomy.org/u/ayshih)\ **Post date:** [November 11, 2023, 3:43am UTC](https://community.openastronomy.org/t/masking-solar-system-objects-in-images/792/2 "2023-11-11T03:43:37Z")
Here are two more pieces of information you’ll need: - How the pixels map to spherical coordinates (in the parlance of WCS, the appropriate projection) - Whether the instrument’s orientation vectors in GCRS are already corrected for stellar aberration (i.e., the shift in the apparent direction of an observed object as a result of observer motion). At high level, here’s how to accomplish what you want for the Earth disc (and the Moon is just the same\*). - Use `get_body()` to get the location of the Earth in GCRS. Ideally you’d specify the accurate observer location and velocity. - Subtract the observer location from the location of the instrument to get the observer-Earth vector. - Take the arcsine of the ratio of the Earth radius and the observer-Earth distance to get the apparent angular size of the Earth. - For every pixel in your data array, calculate the corresponding 2D direction in GCRS for the center of that pixel. This is straightforward if you are working with a constructed WCS. (Of course, adjust your approach if you need something more sophisticated than masking based on pixel centers.) - For every pixel in your data array, calculate the angle between the above 2D direction and the observer-Earth vector (hint: take the dot product) - The mask you want is simply where the above angle array is less than the apparent angular size of the Earth. \* I’ll point out that the position of the Moon can be appreciably inaccurate with Astropy’s built-in ephemeris. I highly recommend you use a JPL ephemeris for Moon calculations. See [Solar System Ephemerides — Astropy v5.3.3](https://docs.astropy.org/en/stable/coordinates/solarsystem.html) ---
**Author:** ![evidlo](https://dub1.discourse-cdn.com/flex005/user_avatar/community.openastronomy.org/evidlo/32/445_2.png) [@evidlo](https://community.openastronomy.org/u/evidlo)\ **Post date:** [November 14, 2023, 10:12pm UTC](https://community.openastronomy.org/t/masking-solar-system-objects-in-images/792/3 "2023-11-14T22:12:52Z")
Thanks for the response! I’ve got working mask code now. For now I’ve avoided `astropy.wcs` since I had some trouble understanding all the [WCS parameters](https://danmoser.github.io/notes/gai_fits-imgs.html), but I think the projection I want is TAN/gnomic. Is there any particular reason to construct a `WCS` object, as opposed to constructing the lines of sight in GCRS directly? ---
**Author:** ![ayshih](https://dub1.discourse-cdn.com/flex005/user_avatar/community.openastronomy.org/ayshih/32/84_2.png) [@ayshih](https://community.openastronomy.org/u/ayshih)\ **Post date:** [November 15, 2023, 4:09am UTC](https://community.openastronomy.org/t/masking-solar-system-objects-in-images/792/4 "2023-11-15T04:09:41Z")
Yup, TAN (gnomonic) is common for imagery. Constructing a WCS header means that you can have wcslib compute the pixel-\>world conversion, and thus avoid re-implementing the transformation code and potentially introducing errors. For the gnomonic projection, a common simplification that people do, sometimes unintentionally, is to use the small-angle approximation, and inaccuracies can creep in. Besides, having a WCS header means that you can do stuff like use the `reproject` package to reproject your image into other projections.