Unsurveyed/Unsourced and other useful linting information

Is there any equivalent of the maplint overlay in potlatch, or the no_name map layer that will show things like ways without a source attribute, or ways which do not appear to follow uploaded GPS tracks?

I’ve come across a lot of stuff created from third party image data, such as landsat imagery (low quality), NPE (low quality) maps, OS streetview (high quality), or Cadastre (high quality) which could do with being verified via local GPS survey. Currently, the only way I can think of to show traced data is to load a region in JOSM, and overlay the data layer with the associated GPS tracks in some bright, contrasty colour. I don’t know of any useful way to display unsourced ways or points without writing a suitable bit of software to transform some OSM data, or rerender sections of the map with rules to highlight data without source tags.

Have I missed some useful application or service which can do this? Or should I start learning the API and putting together some useful tools to do this?

You could create a JOSM style to highlight things with different source tags in different colours etc. Its not too hard to do, just copy and modify one of the examples. Though I’m not sure if its possible to highlight things without a particular tag. http://josm.openstreetmap.de/wiki/Styles

Also I’m not sure how useful this would be. I think its more useful to put the source tags on the changeset, instead of all of the objects. And most things are from a combination of sources anyway. Though it would be useful to highlight things from low quality sources (NPE, Landsat) that haven’t been verified by survey.

I don’t know of anyway of comparing with uploaded GPS tracks, and I’m not sure how useful it would be. There’s no way of knowing how accurate the GPS tracks are, or whether they were actually on the road or on a path next to it etc.

Also, have checked the list of quality assurance tools: http://wiki.openstreetmap.org/wiki/Quality_Assurance
There’s a variety of useful things there, but I don’t know if any do just what you want.

I’ll have a look into that one. If it isn’t possible to do quite what I want, I’m not adverse to rummaging through the source to see how easy the change would be.

Highlighting low quality sources or absent sources is a pretty good start from my point of view. As source tags tend to be attached directly to objects, that’s the approach I’m interested in… I wasn’t aware that additional metadata could be attached to changesets, or even how one would go about reading it. Re-tracing highways and waterways created from NPE using the new OSSV data is a useful task for armchair mappers in the absense of an on-site survey, and for now I just have to scroll around in potlatch or whatever looking for large differences between OSSV features and OSM ways, and taking it from there.

Calculating quality metrics based on node distance from nearest gps point or more nuanced mechanisms such as the mean and standard deviation of distances from gps tracks at 20m intervals along an arc are sufficient. Shading high-scoring nodes/arcs/complete ways red and low scoring ones green is enough of a hint to editors to see where problems might be.

  • Ways built from a GPS track taken along one lane of a multilane road, or a footpath/cyclepath alongside a road would exhibit a small mean offset from the nearest GPS track, but a low standard deviation because they remain parallel to the track.
  • Ways built by tracing GPS tracks at too low a zoom level will exhibit higher standard deviations, because the tracer did not zoom in close enough to see and follow all the wiggliness of the track.
  • Ways built by tracing sattelite imagery will have high mean offsets and standard deviations because they are a long way from any GPS tracks.
  • Ways built by tracing high quality raster images such as Cadastre or OSSV will also score highly depite being higher quality; I’m of the opinion that such ways should be surveyed manually anyway to verify their continued existence, access and highway types, etc.

The thresholds for poor fit could be adjusted to be quite large; half kilometre thresholds for mean offset would ensure that even quite poorly traced tracks, or poor track-to-reality correspondence (such as tracks taken in dense urban environments) would be considered ‘good enough’ and only routes with no surveying or poor surveying would be highlighted. Many such routes exist, even in areas with generally very high coverage and high quality mapping like where I am (Cambridge, UK) and I’d happily plan some bike rides to survey such roads and tracks if only I had an easy way to see where they were!

Poor quality GPS tracks are an unfortunate fact of life on OSM, and there isn’t a whole lot to be done about them. It is unfortunate that there doesn’t seem to be any way to attach any sort of quality tags or other metadata to existing tracks… some areas show a lot of ‘pollution’ by lousy GPS tracks, and it would be nice to be able to draw such things in different colours where it was obvious they didn’t agree with other tracks so they don’t muddy the editor view so much.

Looks like almost all the functionality in those is covered by the JOSM validator plugin which seems pretty comprehensive. Its the poor (or missing) surveying and

…and it looks like I got cut off, and I can’t quite remember what I said here :wink:

But yeah: tools relating to GPS tracks seem to be basically non-existent. Might have to think about filling that particular gap during my Copious Free Time.