How are people handling possible errors with node placement in OSM?

How are people handling possible errors with node placement in OSM? Do you change the node placement in OSM to match GPS track of your run, and how do you accomplish this? The reason I ask is that I am leary about making changes to node placement for the follow reason: In the first photo you can see I have run the following trail many times but the GPS track from my watch seems to be off on a handful of runs so I am not so certain that I can trust the GPS track on a single run. In the second photo, you can see I have run out & back on a non-maintained street and the GPS tracks align closely, but does not hit some of the nodes in OSM. Would you normally move the nodes in OSM?

Screen Shot 2019-12-21 at 1.43.05 PM

This felt like it could stand on its own, so I moved it into a new post.

I think the OSM point of view is that the nodes should align with what’s really out there in the world. I’m going to pass along the question to the Slack group ( OpenStreetMap Chat ) when I get a chance - I think they’ll be able to better articulate the correct response. :smile:

2 Likes

Yes, I agree, it should align with the real world. In most cases we can see it clearly in satellite view and in others, like this case, we cannot see it since it is an unmaintained Class VI road hidden under the tree line. It would be interesting to see how we can determine if the nodes are in the right place and, if not, how we can determine the correct location. I was thinking if we could somehow overlay the GPS track over the OSM map and place the nodes over that track, it would be accurate enough. Thanks James, Let me know if you get any good responses.

1 Like

I’ve seen some conversations about people going out with GPS devices specifically to track their route with OSM:

https://wiki.openstreetmap.org/wiki/Recording_GPS_tracks
https://www.openstreetmap.org/traces

I think that reinforces your question, asking about the best ways to correct for GPS jitter.

I posted in their General channel - if you’re in the Slack group, this link might work for you:
https://osmus.slack.com/archives/C029HV951/p1576965656023000

Here’s a copy/paste of the responses, for those that aren’t in the Slack group:

ian:
Use your best judgment between imagery, GPS traces, and common sense. All of these measurements have some error, so it’s going to be wrong – just make it as good as you can.

phidauex:
Hi, welcome! Citystrides looks like fun, I wasn’t aware of it but now maybe I’ll give it a try.

I’d say there are a few things to consider when adjusting the alignment of a road or adding a new road or path. In order of importance to the map:
That the different ways connect to the right start and end points, and have the right branches. If this is true, then the geometry can be pretty far off and things will ultimately still work out.
That the overall length and tagging (description) of the ways is accurate.
That the way nodes follow the physical landscape as closely as possible.

When working on #3, generally the imagery should “win” - you make sure the imagery is aligned, but adjusting it to match the other major roads in the area. In most parts of the US, the “Bing” imagery is the “primary” imagery on which others are aligned, but that may not be true. Assume for the moment that your local mappers have worked out a good alignment for major roads, and adjust the imagery you are using to match that.

In the case that your user notes where the imagery doesn’t show the path well (hidden by trees), then GPS is a good solution - the traces they show seem generally pretty good, and probably about as good as you will get from any commercial GPS.

There are a few main sources of distortion in GPS traces - the most common is just an offset, which is usually caused by a poor fix, or temporary multipath distortion in an area. In the user’s example, the two “offset” tracks were probably just on a bad day - you can see the offset is consistent through the whole area. If you get 2 or 3 tracks that are well overlaid on separate days, that is pretty darn good for “truth”.

Ideally you’d take one of the GPX files from a day that agreed with the other days (or agrees with other objects in the area that are already mapped), and upload it to OSM using the Traces feature (you can also load a GPX file directly into the iD editor using the “Map Data->Custom Map Data” feature). Use that to trace your own path, just using your eyes and brain to “average” the jitter out. There are some analytical ways to smooth the path, but usually just drawing a nice smooth path where your memories and brain are telling you it should land will do the trick.

Rarely should you actually import the GPX file directly - you’ll either have a path that is far too jittery to use, or you’ll spend more time cleaning it up than you would have spent just tracing a clean path over the line.

In the case of Old Mooar Hill Rd, the road track does seem to be bumped a bit to the east compared to the GPS traces in the area (and looking at Strava heatmap). If the user feels good about their trace, and in particular if they get a second trace that is similarly aligned, then it is totally appropriate to adjust the road position.

One last tip is to always check other imagery, some imagery companies make a special effort to collect “leaf off” imagery in the winter or spring, which makes paths easier to see. In this case, Esri Clarity, while fairly old imagery, actually shows the road pretty well since it is a leaf-off pass. That should help correct the road alignment.

Sorry for the manuscript, hope it helps!

wwhide:
You could always check again the Strava heat map which likely has more gps data. You can add a custom base map in iD using this WMS address https://anygis.ru/api/v1/Tracks_Strava_All/{x}/{y}/{z}

Got that from this page https://github.com/nnngrach/strava_auto_auth/blob/master/README.md

Thank you James. That is a lot of help. Esri Clarity does provide a much clearer satellite view with leaves off and you can see the road much better.

Satellite view is good to use, but also a owned gps trace van really help adjusting OSM to the real world.
When you have an OSM account, you can upload with ease. check this out:

And traces then can be used to verify your adjustments. Especially when you have several traces on the same part that makes it acurate enough to use.

That was my plan originally (to use my GPS track since it seems to match what I saw in Esri Clarity), but I am having trouble locating the activity where I ran that street. I ran it sometime in the last two to three years. I have been more careful over the last year to include the town name in my Strava activity so I can narrow down the search when I am looking for a particular activity. Is there a way to search for the activity, or day the street was run in CityStrides?

Thanks for the info provided, Petje.

i poster this somewhere else: here you can dowload all tracks from your history and see the tracks .You then can klick or hover over each track and find out when that was

petje. I was able to find the activity using the link you provide. Thank you for your help.

1 Like