The logistics of this request are mostly easy… Given a frame representative of AF, draw a box around the area you want to use for AF, have SGPro yell at you if it hates what you’ve done for some reason. For single target sequences, this is great and will work without issue.
For multiple targets, I have been thinking about this for a while now and I am not 100% sure of the best way to do this… especially for targets that are not even visible at sequence start. How does SGPro get AF ROI information for that target? The only way I can think to do this is to essentially pause the sequence and send an alert via the notification system that SGPro is waiting for your AF ROI input. Of course we can save this info so SGPro won’t bug you over multiple nights, but the first part still bothers me. Introducing any step into an automation application that stops the sequence and waits for the user to do something (especially when the requested input is nowhere near sequence start) is of some concern.
An auto focus ROI would be a great feature. I think a target would need to be flagged as using ROI and the imager would need to supply the info prior to the target starting. I suppose that info could be collected in advance for each target in the sequence in the same way the Framing and Mosaic Wizard attaches a copy of the target image. In fact, ROI could be a part of the FMW extension, in that not only are the FMW frames drawn on the image but the ROI box as well.
An extension would be to flag the ROI as inclusive or exclusive. That is, only use the stars inside the box or only use the stars outside the box. If I have heavy nebulosity in the center of the frame, draw a box around it and flag the ROI as exclusive. Then only stars outside the nebulosity get considered for focusing.
Clearly the imager is responsible for managing the ROI info. Changing the rotator position or the “centering point”, could / would invalidate an existing ROI.
Part of the design issue is “robotic imaging” versus “remote imaging” – remote imaging does not mean unattended imaging and asking for ROI information prior to the start of each target is a reasonable first implementation. Perhaps keep it simple and add features as they are identified.
I like Charlie’s idea of integrating it with an image from the FMW. That does not do anything for users who do not have the FMW add-on of course. But as long as you have at least one sub from the sequence, you’ve got what you need to define the ROI.
If you do decide to have a dialog pop up, I think it would be good for there to be a timeout and let the dialog dismiss itself after the timeout. Better to keep going without the ROI than block the sequence indefinitely. I’m imagining the case where you simply forget to specify the ROI then go off to sleep. If the dialog closes before you have a chance to address it (you got the notification but did not make it over to the PC in time), you can always pause the sequence and define the ROI at that time using a FF exposure.
I thought that the ROI features was targeting optical issues, for instance if my telescope is not well corrected I want to skip the corners etc. This is predictable, the offending areas are always well defined. No meridian flip logic needs to be implemented.
If you are trying to address issues related to the framed objects IMO that requieres human interaction at capture time, if you want to select the best ROI to focus you need to test it, you can do some a priori selection but you need to test it at capture time. This feature needs to handle meridian flips.
IMO the optical aberration issues are better handled with simple ROI logic, on the other hand the issues related to particular framed object are better address with good rejection algorithms (automatic or tunable) rather than ROI.
Isolating a Globular Cluster to be NOT included could be as simple as selecting a circular tool, click on the center of the globular and drag your mouse until the circle gets big enough to encompass the globular…
While I agree this is ideal, it is also probably unreasonable. The sheer number of things that can be in a frame make this very difficult (especially to do it quickly on machines that are not super current). But to your point, in terms of DSOs, the existing tuning mechanisms handle probably about 95% of cases so we could say this new feature is gear related and needs no input based on DSO (though I’d be fibbing if I said we had not received requests in equal amount for both of these use cases).
Anyhow… I think the first release of this will not block the sequence in any way and will focus on optical aberations of the imaging train (and therefore not require user input during sequence execution).
Also… if you only had one target, there would be nothing stopping you your from using the AF ROI for DSOs… we just would not support this over multiple targets (initially?).
Personally I would use this for optical problems i.e I want to ignore a central area so focusing is based further out
improving overall focus, when your image plane is not truly flat ( for all sorts of reasons) and this could be truly automatic .
Of course a desired area of focusing to avoid something like a glob is also desirable but can see no way this can be done automatically .