Best APP metric to eliminate sat/plane trails?

@safinsd I am not sure if there is a metric that you can check. However, in tab 6 you should set the "integrate" drop down to either "median" or "average", enable "local normalization rejection" and choose either "winsorized rejection" or "adaptive rejection". In both cases the choice depends on the amount of frames you have. See the mouse hover tool tips for more info.

Once that's all set, you'll need to play with the "kappa low" and "kappa high" values. The lower either is set, the more data gets rejected. Note, as is explained in the mouse hover tool tips, that setting those values lower will increasingly reject good data as well so make sure not to set them too low. You'll need to find a balance between rejecting the trail and rejecting signal that you actually want to appear in the integration result.

What is kappa - i.e. what levels of probability does it correspond to (e.g. sigma score of 3 corresponds to 99%). For a very low number of frames (e.g. 2) and with two types of outliers - plane trails (light) and trees (dark) what would be optimum values of kappa? Plane trails are VERY bright and tree branches are VERY dark so they should be easy to detect.

@ddnum See https://www.astropixelprocessor.com/community/main-forum/meaning-of-kappa-low-and-high-for-winsorized-average/

@wvreeven 

Thanks much for your very helpful reply. I selected median-average and local normalization rejection-adaptive rejection. Looking forward to the result.

Thanks for the link where Mabula said "Kappa is the number of times the scale/dispersion factor is applied for rejection. Scale/dispersion is usually the standard deviaton." I knew that but I was hoping for a translation into probability because "number of times" has no meaning to me. Is kappa like a z score?

@ddnum, kappa is simply a number which you set to control how many times the scale/dispersion is used to determine which pixels should be rejected or not.

For example a high kappa of 3, means that all pixels above

central value + 3*scale

will be rejected in all pixelstacks of your integration.

If the pixelstacks would have a nornal/gaussian distribution, then the following applies: https://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule

Unfortunately, the pixelstacks only approach a gaussian, so the probability is slightly different and lower as a consequence...

If APP's automatic rejection still leaves outliers like airplanes, you want to disable automatic integration and control the rejection settings yourself like Wouter @wvreeven indicates 😉

The settings that APP used in the automatic integration are stored in the FITS header of the integration, so you lower the kappa high and low from those values until all is to your satisfaction.

Mabula