[talk-au] Well I bought an L1/L5 Mouse GPS

[stevea]

On Jun 22, 2022, at 6:45 PM, Alex Sims <alex at softgrow.com> wrote:
> I’ve now got a relatively (<$100 + postage) Mouse GPS. It is amazingly accurate. That’s the good news.
>  
> Now I can see a whole bunch of streets, buildings etc out by 1-5 meters as *some* features were traced without correcting the image offset. Also found my cheap GPS and an OSX machine are not a great combination. GPSD gets confused and ends up with mojibake most times. Virtual serial port is fine in screen, QGIS etc.
>  
> Still fun

I don't know the specific product, I'll look it up (and I'm in the States, maybe I can't get one, or only with difficulty, not worth shipping / taxes / duty / customs, whatever).

Image offset-derived "smears" in OSM's data can be a pain in the neck, especially as you get "multiple drifts" from a number of (imaging) sources.  However, 1 to 5 meters, today (2022) isn't terrible.  And over years (yeah, sorry for those in a hurry), I have noticed "this gets better."  It's an ongoing effort.  Think of it like a slow motion countdown to zero, but like Zeno's Paradox, never really quite gets there.  Too, precision can be in the eye of the beholder.  1 to 5 meters?  Hey, you're not lost, perhaps a bit "smeary" but not lost.

JOSM does have a way to "imagery offset" differing layers, though I find it's only a temporary solution and try to keep these at 0,0.  My recommendation is to not try to do too much correcting all at once.  Chip away at it.  The data do get better.  Slowly, surely.