[Talk-GB] Gathering street level imagery off road?

Chris Hodges chris at c-hodges.co.uk
Thu Jun 24 12:49:46 UTC 2021 

That's a nice, and interesting, way to sync times.

I've lost a few phone mounts to potholes, once catching the phone and 
half the mount in midair but more often damaging something.  Some of the 
apparently metal designs have a piece of plastic at a critical point 
(like a ball joint made of plastic, that a screw threads into) and I've 
once snapped the metal screw holding things together.  I do use heavy 
phones though, which are also too fat for some mounts.

Do you mean Mapillary automatically works out the direction from the 
stream of points, during upload? That would certainly make things easier 
- and if you're on Windows their tools sound good.

Can I ask though, why you went for video rather than stills? Is it just 
a matter of what your cameras offer?  I get much better image quality on 
mine with stills, and can get 1 per second which is plenty.  A fairly 
large  SD card lasts longer than the battery on my ActiveOn CX, which 
does about 3 hours; I've hacked a case to power it off my dynamo or an 
external battery but still have to sort out a modified mount. Perhaps 
yours do better video than mine

Chris



On 24/06/2021 13:22, Jon Pennycook wrote:
> I take action camera footage and upload it to Mapillary. I used to 
> have a mobile phone mount but it snapped after going over a pothole - 
> avoid plastic!
>
> I record footage using my Cycliq Fly12 CE handlebar camera (I have 
> also used OLFI ONE.FIVE Black, Crosstour CT9900, and Cycliq Fly6 CE 
> GEN 2 - two helmet cameras and one rear camera - from time to time). 
> For GPS, I use Locus Map set to record every 2 yards or 0 seconds, 
> recording only when moving. To sync the camera and GPS, I record 10s 
> of https://time.gov/ <https://time.gov/> whenever I start recording 
> and periodically (every 2 hours or so) after that - the UTC time is 
> what I need. Direction is not important as it will be interpolated 
> between two images.
>
> Finally, I use a set of PowerShell scripts that I wrote to crop the 
> timestamp from the videos, then convert them to geotagged images, and 
> finally upload them to Mapillary (the scripts use exiftool, ffmpeg, 
> and mapillary-tools). Before uploading, I check that the images have 
> the correct position (if not, I tweak the offset between image time 
> and GPS) and delete rubbish images (including footage of me recording 
> my phone showing time.gov <http://time.gov>)
>
> Scripts here: https://gitlab.com/jpennycook/bikecam 
> <https://gitlab.com/jpennycook/bikecam>
> I use PowerShell 7 on Windows 10, but I expect it would work with 
> PowerShell 5.
>
> Jon
>
>
>
>
> On Thu, 24 Jun 2021, 11:33 Chris Hodges, <chris at c-hodges.co.uk 
> <mailto:chris at c-hodges.co.uk>> wrote:
>
>     I've seen a high mounted setup for a bike but really wouldn't
>     fancy it:
>     https://www.openstreetmap.org/user/tordans/diary/395215
>     <https://www.openstreetmap.org/user/tordans/diary/395215> (probably
>     spotted via the weekly OSM newsletter).  That's for a GoPro with GPS.
>
>     Personally on a bike I'd go for a standard handlebar mount - there
>     are
>     some cheap all-metal ones on ebay that don't block the camera.  I
>     use a
>     waterproof phone with a lanyard attached to the bike as well, as I've
>     killed a phone by dropping it off the bars too many times. Then
>     either
>     mapillary/KartaView's own app or timelapse software with geotagging
>     (e.g. rolling my own using Tasker, which is running a script as I
>     ride
>     anyway).  I've looked into doing this myself but I navigate on my
>     phone
>     when I go anywhere interesting and the angle I need to see the
>     screen is
>     wrong for taking pictures.  I don't currently have a spare phone
>     to test
>     otherwise I might try again.
>
>     The other option I considered is using timelapse mode on a
>     handlebar-mounted action camera, synchronising its clock to my phone,
>     and using the image timestamp to extract the coordinates from a
>     GPX file
>     recorded by the phone.  Mapillary says there are tools to combine the
>     data sources
>     https://help.mapillary.com/hc/en-us/articles/115001717829-Geotagging-images
>     <https://help.mapillary.com/hc/en-us/articles/115001717829-Geotagging-images>
>
>     but anyway it wouldn't be hard using Python with libraries to
>     handle the
>     EXIF data and parse the GPX.
>
>     One slight downside is 1s resolution but even going downhill on a
>     bike
>     that's not going to be more than about 10-15m so only a little worse
>     than the GPS error.  Another is that geotagging for this purpose
>     needs
>     more than just a lat/long; it needs a direction (even if you have
>     a 360°
>     camera). That means more than simply grabbing a set of coordinates at
>     the same time as the image was taken, instead using the neighbouring
>     coordinates to figure out the direction of travel.  This then breaks
>     down if you're travelling too slowly, when the random component of
>     the
>     GPS error becomes comparable to the distance between points (some
>     of the
>     error is correlated between fixes so it's not too bad, and you could
>     always find the tangent to your track based on more points).  I'm not
>     sure how sophisticated the tools linked by mapillary are for this.
>
>     Vague ramblings I know, but hopefullly there are some useful
>     hints, even
>     if only at those links
>
>     Chris
>
>
>
>     On 24/06/2021 10:50, Mat Attlee wrote:
>     > What's best practice for gathering street imagery off road ie foot
>     > paths, forest trails etc? I've started using Karta View to capture
>     > imagery during my run but wondered if there was a better way of
>     doing
>     > things.
>     >
>     > Also has anyone ever had much success using Karta View or
>     Mapillary on
>     > a bicycle mount - that is assuming such things exist?
>     >
>     > _______________________________________________
>     > Talk-GB mailing list
>     > Talk-GB at openstreetmap.org <mailto:Talk-GB at openstreetmap.org>
>     > https://lists.openstreetmap.org/listinfo/talk-gb
>     <https://lists.openstreetmap.org/listinfo/talk-gb>
>
>
>
>     _______________________________________________
>     Talk-GB mailing list
>     Talk-GB at openstreetmap.org <mailto:Talk-GB at openstreetmap.org>
>     https://lists.openstreetmap.org/listinfo/talk-gb
>     <https://lists.openstreetmap.org/listinfo/talk-gb>
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb

More information about the Talk-GB mailing list