Faint comet in rich star field - manual registration mode?

Did you already try the comet registration tool in APP to see if it detects the comet properly? Normally it does, there was some improvements done on detecting nuclei as well, it sometimes will ask to manually select it again if it proves difficult.

@vincent-mod 
I have used the comet registration tool for all my recent comet images. I am giving APP a difficult problem because I'm imaging a faint comet against a rich star field. Here is one example. The crosshairs are actually not on the comet. The comet is several pixels above the crosshairs.

For that image I was unable to manually coax the crosshairs onto to comet, so I did not use it in the final stack. Images from several minutes later, when the comet had moved farther away from the distracting star, I was able to manually place crosshairs on the comet...but it required manual intervention on the vast majority of images. 

Here is my most recent image of comet 104P Kowal.

This was a stack of 236 images. Because of the rich star field, I had to manually position the crosshairs on at least 3/4 of the images. The centroiding routine appears to be quite accurate, even on this faint comet, but the requirement for so much manual crosshair positioning is tedious.

Ah I see, that is indeed a bit challenging. Would be great if you could share this data with us, I'll also forward your request to Mabula. If you can, please upload to the server mentioned top right of the forum, in a directory called "tomkrajci-manualRegMode".

@vincent-mod Uploaded in several zip files.

Note that I have separated the light frames into the three sessions of the barn door drive.

The barn door drive is not perfect. In the middle of each run tracking performance/image quality is pretty good and pretty consistent. Near the beginning and end of each run fewer images are of acceptable quality. If you want to save time while evaluating registration algorithms, select only a small subset of images from the middle of each of the three runs.

When I processed the images I opened the lights in 'super pixel' mode. I wanted to go deeper with a 2x2 bin, and I think the improvement in image resolution using 1x1 binning is not very large because of the tracking/image quality.

I hope this set of images and calibration frames helps.

Posted by: @tomkrajci

Here is my most recent image of comet 104P Kowal.

This was a stack of 236 images. Because of the rich star field, I had to manually position the crosshairs on at least 3/4 of the images. The centroiding routine appears to be quite accurate, even on this faint comet, but the requirement for so much manual crosshair positioning is tedious.

Hi Tom @tomkrajci,

Thank you very much for sharing your data.

Looking at the result, it seems APP worked fine allthough it took a bit of manual labour for you. If the comet is weak like this, if the current comet detection has low certainty, it will ask you to manually select the core. If stars are nearby of similar or higher brightness, then the comet detection might fail yes. But since APP is calculation comet speed relative to the stars, it can correct those images as you can see in the final result.

I have added to our issue list to test your data to see if we can improve things further 😉

Mabula