NavList:

Sure---there are a few gotchas with the tools that I've straightened out only recently after some trial and error.  For example, I use astrometry.net to get an initial position for the star field.  Do not have it do a distortion estimate!  If one is included, SCAMP will get confused.  The "--no-tweak" option does the trick.  A few other things like that.

If you haven't tried it yet, drop a few star snapshots into nova.astrometry.net, the web service (no software download needed).  It will figure out the photo's position, scale, and rotation, and give nice annotated renderings of your photo with star names, etc.  It doesn't do full weighted-least-squares fitting across many sources, though; use SCAMP for that, after you've dialed in a first guess with astrometry.net (which is usually pretty good, within 10 or 20 arcseconds).

 

That is challenging because of the very small aperture of the cellphone cameras.  The sensor itself is quite good these days (back-illuminated, microlenses, low read noise), but it cannot overcome the small lens.  Also the field of view is typically fixed and very wide.  Even so, the wide field is likely to include brighter stars, so that's a plus, and the effective angular resolution of the system is dependent on both pixel size and source SNR.  I don't know where the best tradeoff lies for this application.

Point-and-shoot cameras might be a happy medium here.

Cheers,
Peter

ps:  maybe this has come up before: surely cameras could help with sextant observations.  (I've tried a few snapshots through the sextant but nothing within the observation loop.)  Imagine the usual daytime sextant sight.  The Sun is brought down close to the horizon, then the sextant is swung.  If a camera (or cellphone) is mounted on the sextant in place of the user's eye, then there's no need for tangency---software can easily estimate the small angle from limb to horizon, then add it to the sextant setting.  Timestamps are also accurate and automated.  And, three photos is enough to estimate the arc as the sextant is swung, since three points make a circle (although the observations are not simultaneous, but that's something else the software can easily handle).  Just a thought.  Performance is limited more by the quality of the horizon and the refraction model, though.