Hello Lyuboslava,

As you said, the first thing that came to mind was running a CV program to read the needed data. If the data does not change in the duration of the footage, then, in my humble opinion, it would be easiest to take a single frame from each clip and run an optical character recognition program to get the temperature.

The moon phase might pose a bit more of a challenge if the meta data truly is hidden. A different CNN program that is trained on the moon phases should suffice. But as you well know, not the simplest solution.

It's certainly possible that your camera(s) just aren't embedded structured metadata for this information.  But I wouldn't give up too quickly on looking for it, because extracting it from actual video frames is more challenging, less efficient, and less reliable.

EXIF / IPTC metadata is effectively limited to a finite set of pre-defined attributes, mostly centred around camera settings & author information, respectively.  You can check the latest specification to see precisely what is formally defined.  Ambient temperature actually is in there (and has been for a while) but there's nothing for astronomical data (although, can you not infer the moon phase based on the location & time data?).

It is alas a long-running issue that camera makers often misunderstand the specification and store data incorrectly (or not at all).

Camera makers can embed proprietary data in MakerNote fields, but the format is unspecified and standard tools (like exiftool) likely won't understand them for all but the most popular cameras (think Nikon, Canon, Sony, etc).  And possibly only for older models.

Many common image & video containers (e.g. JPEG and MP4) do support embeded XMP, an extensible format which allows model- and make-specific metadata.  It's possible that's in use in your case.  exiftool does not show embedded XMP data by default - at least not reliably.  In a nutshell, I suggest running exiftool as:

exiftool -D -g -G -ee -scanForXMP -u -U <PATH>

That's more likely to show you embedded XMP data, as well as all EXIF / IPTC tags (even those unrecognised by exiftool - you can at least get their raw bytes and work from there).  You can also add the `-v3` argument if you want to dig even harder, or see the raw data in more detail.

Another way some cameras store proprietary metadata is in proprietary tracks (e.g. GPMF for GoPros).  You can see if your files contain any such tracks via:

ffmpeg -i <PATH>

Usually these are used for metadata which can change during a recording, as EXIF / IPTC / XMP have no real support for that.  But some camera makers use them even for static metadata out of e.g. laziness or hubris.

Extracting such tracks is easy enough with tools like ffmpeg, but interpreting them can be challenging if their format is not documented like GoPro's GPMF.

Can you provide a sample file for us to examine?

Hi Tobias! 

This sounds great and I am looking forward to trying it out after returning from field work! 

Very cool with the Vectronic Activity data! I am looking forward to check How we can use that! 

Cheers,

Lars