Alright… just a check in here. I did some work to expose this metric. The goal, in the is case, is to see if we can provide something useful in a pretty quick way so that it doesn’t compete with AutoFlats and Custom Horizons (SGPro 4.3). Assuming this is somewhat useful, it can also go into 4.3, but before we talk about that, let’s review the state of things.
In order to get an accurate measure of eccentricity, I was forced to re-evaluate the way in which we find and measure star fields. The current method in SGPro uses a method that is a compromise between accuracy and speed. It really just deals with centroid calculations by bounding the star in a rectangle and then finding the centroid in said rectangle by calculating the center of gravity as dictated by pixel intensity. All said, it’s fairly accurate and reasonably fast. BUT… this method does not produce accurate measurements for star eccentricity. The work I have completed recently refactors this method a good deal and uses proper ellipse detection through a least squares fit. Much to my surprise I was able to tune this method such that the find stars method only incurs an additional time penalty of 1 ms per star on the image (usually around 1/3 of a sec for most images). I think this means that eccentricity can always be measured, but SGPro will still provide an option to disable it… the reason being that SGPro is already a wall of data and stats. If you don’t use or need eccentricity, you don’t need to have it add to the clutter.
So now we have access to eccentricity, but how does it become useful in a way that does not require weeks of work. This is my proposal, but I want to see if it meets needs regarding the subject of this thread, if it falls short and if there are any quick adjustments that would close the gap between those 2 things.
An in-depth look at this is here:
Specific to this thread, and as an example, from the “bad data” zip file you provided us.
Image:
BAD_NGC2070_O_600sec_gain_100_offset_50_20221102_040436_022_fpos_17340_-10C_amb_12.4C.fit
Eccentricity: 0.5
Aspect Ratio: 20%
This doesn’t really mean anything though… What I am hoping you might provide is a sample of a few images from the same events (filters, gains, etc) that you have classified as “good”. In my mind, the absolute number isn’t the most important aspect of this measurement, but rather that there is an actual difference in the measurements between good and bad.