[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Automatic Processing of Data


Arne

I would like to take a stab at automating the subjective very bright star
detection and subsequent rejection for flat fielding.  Would you care to try
to define how you believe you do this (loosely)?  Do you have a data set you
have done this process for which I could get a copy of?

Do you (or any one else) think that a lower limit on the number of frames to
combine to generate a decent dark/flat-field image?  The reason I ask is I
have some software which does the following:

Starting will means of all images, and clipping values of 0x0000 to 0xFFFF
(i.e. no clipping first pass).
1	Calculate the mean and sigma of data
2	Calculate new clipping values based on the mean and sigma (mean +=
X*sigma)
3	Run through list of values, keeping data within clipping range
4	If any data was removed, go to 1, otherwise, use images not rejected

Right now, I've used X as 2.5 and 1.5 with good success for removing images
with large means (twilight shots).  When I drop X down to 1, for one set of
24 dark images, only 2 make it through this iterative process.  I'm going to
modify the above algorithm so that X is adjusted from 1 to some max value,
incrementing X when the number of images is less than some critical value
and/or % of available images.  The number of dark frames in the data sets I
have varies from 24 to 80...  Do you even thing that this is a valid way of
picking images to use (for dark and flat fielding)?

Thanks,
Rob
 

>   As Mike said, each sky frame used to make the median should
> be scaled by its mode or median.  I use mode.  I throw out
> any frame which includes a very bright star (subjective, usually
> if it bleeds a lot).  This manual operation, of looking at all
> data frames to hand-pick the set for making a flat, is the
> time consuming process of reducing a night of data.

>   Dark current should be relatively constant at a constant
> temperature.  So if you are really regulating the TEC at, say,
> +/- 0.1C, then I might trust darks from night to night.  However,
> remember that they are low level, and to get decent signal/noise
> so that they don't contribute much noise to the processing, you
> need to median-combine lots, not just 6.  This hurts your faint end
> the most.
> Arne
>