On Tue, Oct 29, 2024 at 11:42:23PM +0000, Adrian via Talk-GB wrote: > >On Friday 25 October 2024 at 16:45:57 GMT+1, ael via Talk-GB > <[email protected]> wrote: > > >Just a comment. I capture gps traces (usually on a garmin handheld) and > also images (normally jpeg) from various cameras including a dashcam and .. [snip].. > way just to upload them. This seems like a glaring omission. My Nextbase > dashcam produces geotagged mp4 video, and there is no simple way to > upload those. I do realise that the gps information in the mp4 videos is > not standardised, so there may be an excuse for the video. Not to > mention the sheer size of the files. > > > A user described how they uploaded a georeferenced video here > https://www.openstreetmap.org/user/Branko%20Kokanovic/diary/403952
Thanks for that link: very interesting but I will need more time to read it all. > I have a Nextbase dashcam and the positions it gives are not very accurate. > They are generally less accurate than you would get from a smartphone. I have > even seen position offsets of up to 30m in a location which is not > particularly difficult. The speed seems to be pretty accurate however. You don't mention which model, but I have a 312GW and I have done extensive checks on its gps quality. I had to do a fair bit of work to find out how to extract the gpx tracks from the video, and wrote a program to do that: https://github.com/clarified/mov2gps.git I also record tracks from a reasonable garmin handheld and a satnav running navit alongside the Nextbase 312GW, and the Garmin tracks are always more accurate. My observations are that both the Nextbase and the satnav are usually fairly good when there is an open sky, but under tree cover or around high buildings, they can struggle and errors like 30m in those circumstances are not uncommon. I suspect the garmin has a much better antenna and also captures more satellites. > I looked closely at a section of footage and found other issues too. The > in-vision speed and position data are slightly delayed relative to the video. Yes. I also measured this and estimated the delay. From memory, it is around 1 second but maybe the 2 seconds that you mentioned. I apply a displacement, which depends on the speed, of course, when working from the Nextbase video. >I had a Garmin hand-held SatNav (GPS receiver) mounted on the windscreen. I >used it to record a tracklog at the same time. It gives considerably more >accurate positions than you would get from a smartphone. It helped me find >these issues. We seem to have been doing very similar things: if I had known about your work, we might have saved some time and effort by working together. > The video file recorded by the dashcam contains a data stream. You can > extract the data stream using the -ee option of Phil Harvey's ExifTool. I think that I found that at some point. My mov2gpx does a bit more in some ways specifically for the Nextbase 312 video. It has a -debug switch which displays a lot of information about the embedded data. > The stream includes the actual data from the dashcam's SatNav chip.d This > data is in the usual NMEA format. The data is not easy to work with because > (a) the checksums have been clobbered and (b) the NMEA data is mixed up with > other data. mov2gpx decodes all of that, at least for the 312GW video. Nextbase did send me some video samples from some of their other models, and I was shocked at how different the video encoding was across their range. I had intended to expand mov2gpx to cover all the options until I saw how much work was involved. You may have a different model with different video encoding. I guess we are getting off-topic: my fault. Perhaps we should start a new thread if there is more to say. ael _______________________________________________ Talk-GB mailing list [email protected] https://lists.openstreetmap.org/listinfo/talk-gb
