finding star-like objects in images

[Helder]

Hi,
I'm not an astronomer and I guess that this is something that astronomers have been confronted with quite often in their lives.
I have a detector where particle events generate intensity across some pixels (2-5 x 2-5) [*]. Typically their integral intensity is constant (lets say 100 +/- 20). These events show over a noisy bkg.
Apart from having a constant intensity, these events are similar to stars (that have a varying luminosity).

What approaches are typically used for detecting/locating such events?

Any IDL solution readily available out there?

Thanks for reading so far and for any suggestions.

Regards,
Helder

[*] - threshold methods would not work very well, because the total intensity of 100 may be distributed over 2x2 pixels (~25 per pixel) or 5x5 (~10 per pixel).

[Markus Schmassmann]

Post by Helder
I'm not an astronomer and I guess that this is something that
astronomers have been confronted with quite often in their lives.
I have a detector where particle events generate intensity across
some pixels (2-5 x 2-5) [*]. Typically their integral intensity is
constant (lets say 100 +/- 20). These events show over a noisy bkg.
Apart from having a constant intensity, these events are similar to
stars (that have a varying luminosity).
What approaches are typically used for detecting/locating such
events?
Any IDL solution readily available out there?
Thanks for reading so far and for any suggestions.
[*] - threshold methods would not work very well, because the total
intensity of 100 may be distributed over 2x2 pixels (~25 per pixel)
or 5x5 (~10 per pixel).

Hi Helder,

just a guess, but have you tried

star=where(smooth(img,[5,5]) gt 100./5^2/2)

you will have to change the threshold value and maybe also how much
smoothing you apply, but it could work.

good luck, Markus

[wlandsman]

You could try find.pro based on a popular software package(DAOPHOT) used by astronomers
https://idlastro.gsfc.nasa.gov/ftp/pro/idlphot/find.pro
The image is convolved with a lowered Gaussian with the approximate FWHM of the stars.

Note, though, that astronomers want to detect stars but not cosmic rays so there are sharpness and roundness criteria (with stars being less sharp and more round than cosmic rays).

--Wayne

Post by Helder
Hi,
I'm not an astronomer and I guess that this is something that astronomers have been confronted with quite often in their lives.
I have a detector where particle events generate intensity across some pixels (2-5 x 2-5) [*]. Typically their integral intensity is constant (lets say 100 +/- 20). These events show over a noisy bkg.
Apart from having a constant intensity, these events are similar to stars (that have a varying luminosity).
What approaches are typically used for detecting/locating such events?
Any IDL solution readily available out there?
Thanks for reading so far and for any suggestions.
Regards,
Helder
[*] - threshold methods would not work very well, because the total intensity of 100 may be distributed over 2x2 pixels (~25 per pixel) or 5x5 (~10 per pixel).

[Helder]

Post by wlandsman
You could try find.pro based on a popular software package(DAOPHOT) used by astronomers
https://idlastro.gsfc.nasa.gov/ftp/pro/idlphot/find.pro
The image is convolved with a lowered Gaussian with the approximate FWHM of the stars.
Note, though, that astronomers want to detect stars but not cosmic rays so there are sharpness and roundness criteria (with stars being less sharp and more round than cosmic rays).
--Wayne

Post by Helder
Hi,
I'm not an astronomer and I guess that this is something that astronomers have been confronted with quite often in their lives.
I have a detector where particle events generate intensity across some pixels (2-5 x 2-5) [*]. Typically their integral intensity is constant (lets say 100 +/- 20). These events show over a noisy bkg.
Apart from having a constant intensity, these events are similar to stars (that have a varying luminosity).
What approaches are typically used for detecting/locating such events?
Any IDL solution readily available out there?
Thanks for reading so far and for any suggestions.
Regards,
Helder
[*] - threshold methods would not work very well, because the total intensity of 100 may be distributed over 2x2 pixels (~25 per pixel) or 5x5 (~10 per pixel).

Dear Markus and Wayne,
thank you very much for your insight. It took me some time implement the above (as part of a bigger analysis) and I'm now very happy with it.
Regards,
Helder