Re: [OSM-talk] 'Distance to feature' maps?
On Tue, 14 Jul 2009 23:51:22 +0200 Frederik Ramm frede...@remote.org wrote: I happen to have a Perl script that does something like this With some pretty simple mods (which even I could handle) your code can output a CSV list of lat/long/distance; however I struggled produce a decent contour from this. Previously I had used a perl script to call Gnuplot to 'contour' the SRTM data, which worked fine. Using the data from your script had the problem that is was 'sparse' point data rather than an XY full table of data and as such Gnuplot didn't do a good job at creating contours. There are 2 ways around this I can think of: 1) split the ways in the source data into many more nodes, so that there is more chance of Gnuplot seeing a line trend. 2) Play some math tricks to interpolate the output data, effectively performing the same thing. Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
simon, i might have a christmas present for you... please have a look at the files below: http://www.gary68.de/temp/hofheim.png http://www.gary68.de/temp/hofheim.svg http://www.gary68.de/temp/hofheim.osm http://www.gary68.de/temp/hofheim.gpx you could insert the *.osm in your original osm file and have a render rule for those entries. then you can use a renderer of your choice, like KOSMOS. Don't know if gpx is useful though. and here's the code (you will need the modules too!) - first BETA. no QA took place :-) http://svn.openstreetmap.org/applications/utils/gary68/distancemap.pl http://svn.openstreetmap.org/applications/utils/gary68/ any more wishes? ps: i don't know if shapes like elevation lines really make sense here... pps: wiki page will follow! cheers gary68 gerhard On Tue, 2009-07-14 at 13:20 -0400, si...@mungewell.org wrote: Has anyone done any work in the area of making maps representing 'distance to feature' as measure along routable roadways/footways? Local school board is trying to drop school bus service school kids (from grade 1 up) within 2.0km of school and get them to walk to school (bear in mind this is a rural community with -40'C winters, wildlife such as bears/cougars and an ungated CPR train route through it). I have some ideas about representing the data, but does anyone have any suggestions on how to parse through each node and measure the distance along each way to produce a list with 'distance to school' tag for every hightway node? Is this the 'best' approach? The node list would/could then be passed through a variant of 'strm2osm.pl' to produce a contour layer, and then merged/plotted with osmarender. I could also imagine that this 'work' might be usefull in 'distance to nearest [bus stop, post box, etc]' maps/datasets. Cheers, Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
simon, i might have a christmas present for you... Is it Christmas already?? I like presents ;-) Looks like we've both been busy (just don't tell my boss I was up till 02:30 playing with this). Using a combination of Fredrick's perl script and GnuPlot (to perform contouring) I produced the attached contours. You should be able to open in JOSM and then download the OSM lay to match. I'll merge '.osm's and throw this through osmarender tonight to see what they really look like. I'm a visual sort of guy, but you might be right them being confussing to the average bod. http://svn.openstreetmap.org/applications/utils/gary68/distancemap.pl One thing I was thinking was that the osm used for producing the distance nodes, does not have to be the same as the mapping layer. This would enable it to be customised (unroutable ways deleted for example) prior to plotting. There are also 'areas' for improvement, for example the main highway running through Coleman is missing some sections of side walks and really kids should only be crossing at signed crosswalks. This could be achieve by duplicating the way slightly displaced and editing to force the above. You might also want to factor in the 'access=private' tag and avoid stairs (in the case of a walking map) as mum dropping kids at school often have prams/strollers for siblings. Cheers, Simon. PS. I started a wiki page to write up our efforts: http://wiki.openstreetmap.org/wiki/Distance_maps ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
On Wed, 15 Jul 2009 09:00:23 +0200
Frederik Ramm frede...@remote.org wrote:
Finally managed to get something 'believable', see attached contours. You can
download the corresponding roads with JOSM.
Problems where:
1) GnuPlot output number format didn't have enough digits and was distorting
data.
2) Gnuplot dgrid3d needed to be tweaked to improve rejection of points further
away.
3) Source data didn't contain enough nodes, had to interpolate
So I need to merge and render this out with Osmarender, which shouldn't be too
difficult. And then write it up in some form for others to try.
I'd also like to look at selecting which ways to route down. At present it
looks like it is fixed in the perl script but there are some it shouldn't use
('access=private') and it would be nice to do two maps, one with stairs and one
without (mother pushing pram...).
Cheers,
Simon.
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
only needs an dijkstra algorithm and some more lines of code. sounds so simple... ;-) I had a look at the wikipedia page (http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm) and that would look to make sense. As you may have noticed I've actually had a relatively amount of success with this, but my brain is off on the 'multiple targets' question. Is it valid to repeat the dijkstra algorithm with multiple 'target' nodes, so you could find the shortest path to (say) any playground? Target nodes could auto-magically be extracted from OSM file based on tags. Is this as simple as keeping the 'found' state in a seperate variable for each node (rather than using the distance variable as a found state)? Then you could 'walk' from each target node looking at 'found' state and 'distance' to determine whether this node was closer to new target than the previous, if it were you adjust the 'found' to indicate which target was closer, set your distance and then continue to walk. You'd also have a record (in the 'found' state) of which target was closer. Does this sound correct? Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
On Wed, 15 Jul 2009 13:30:44 -0400 (EDT) si...@mungewell.org wrote: only needs an dijkstra algorithm and some more lines of code. sounds so simple... ;-) I had a look at the wikipedia page (http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm) and that would look to make sense. Actually, an implementation of the Dijkstra algorithm is not difficult at all. When we were trying to do this with Python and Django, we came across a couple of sample implementations. Better yet, there is the Boost graph library which already implements the Dijkstra shortest path algorithm. The library is in C++, but there are various bindings to it. To push our own work, one can take a look at an implementation at http://citybyroad.com/ (Log in with user/password testuser/ gpstrack). This actually was a more ambitious attempt to get shortest paths based on actual travel time from GPS traces. However, ignore that for the moment, and also ignore the half- completed state of the application. Leave all the drop-down boxes as Unspecified, enter a from/to address via the auto- complete boxes (try Raja Puri and Srijan Technologies, allowing the auto-complete to take over after typing a few letters), and click on the blue FIND ROUTE button. The result shown on the map is from the Dijkstra shortest-path algorithm, based on distance. The distance is calculated by integrating along known routes, which is a fairly tedious, manual process, but OSM routes should already have that information. For someone familiar with Delhi, the application will be unable to find some routes, as the database is not completely populated. As you may have noticed I've actually had a relatively amount of success with this, but my brain is off on the 'multiple targets' question. Is it valid to repeat the dijkstra algorithm with multiple 'target' nodes, so you could find the shortest path to (say) any playground? Target nodes could auto-magically be extracted from OSM file based on tags. [...] Does this sound correct? Your description covers the gist of the issue. The Dijkstra algorithm actually visits every node in the network, and thus running it once between two points should find all shortest paths. In practice, it seems that the Boost implementation prunes unlikely paths, so that one has to start with two extreme points, and maybe rerun the algorithm a couple more times with other extreme points in order to get full coverage. Came into this conversation a little late, but if you had a definite purpose in mind, we would be glad to share the code, and maybe even put in some work on a reimplementation. What language are you planning on using? Regards, Gora ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
of course this is possible. store the new min dist for each node if it is lower than the old min dist. store next postbox node id if wanted. all a matter of time and memory :-) On Wed, 2009-07-15 at 13:30 -0400, si...@mungewell.org wrote: only needs an dijkstra algorithm and some more lines of code. sounds so simple... ;-) I had a look at the wikipedia page (http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm) and that would look to make sense. As you may have noticed I've actually had a relatively amount of success with this, but my brain is off on the 'multiple targets' question. Is it valid to repeat the dijkstra algorithm with multiple 'target' nodes, so you could find the shortest path to (say) any playground? Target nodes could auto-magically be extracted from OSM file based on tags. Is this as simple as keeping the 'found' state in a seperate variable for each node (rather than using the distance variable as a found state)? Then you could 'walk' from each target node looking at 'found' state and 'distance' to determine whether this node was closer to new target than the previous, if it were you adjust the 'found' to indicate which target was closer, set your distance and then continue to walk. You'd also have a record (in the 'found' state) of which target was closer. Does this sound correct? Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
[OSM-talk] 'Distance to feature' maps?
Has anyone done any work in the area of making maps representing 'distance to feature' as measure along routable roadways/footways? Local school board is trying to drop school bus service school kids (from grade 1 up) within 2.0km of school and get them to walk to school (bear in mind this is a rural community with -40'C winters, wildlife such as bears/cougars and an ungated CPR train route through it). I have some ideas about representing the data, but does anyone have any suggestions on how to parse through each node and measure the distance along each way to produce a list with 'distance to school' tag for every hightway node? Is this the 'best' approach? The node list would/could then be passed through a variant of 'strm2osm.pl' to produce a contour layer, and then merged/plotted with osmarender. I could also imagine that this 'work' might be usefull in 'distance to nearest [bus stop, post box, etc]' maps/datasets. Cheers, Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
(bear in mind this is a rural community with -40'C winters, wildlife such as bears/cougars and an ungated CPR train route through it). Of course everyone like as a map link: http://www.openstreetmap.org/?mlat=49.6363mlon=-114.5065zoom=14layers=00B0FTF ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
This is generally the sort of thing done with a full blown GIS. Which is how I would approach it. Use OSM as a data file, stripping out all of the un-needed information, create a buffer around the schools of interest, maybe map registered sex offenders within the buffer (if you have the equivalent where you are) and all known hazards (I'd probably call unlit, unplowed or forest path hazardous), and then create maybe a heatmap of danger. All of this should be doable in something like QGIS. Good luck! -Tyler ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
If you're prepared to take the route it gives you, cyclestreets.net gives you distances along the route (as opposed to how the crow flies), as would any good routing system I'd expect. David ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
On Tue, Jul 14, 2009 at 5:42 PM, Tylertyler.ritc...@gmail.com wrote: This is generally the sort of thing done with a full blown GIS. Which is how I would approach it. Use OSM as a data file, stripping out all of the un-needed information, create a buffer around the schools of interest, maybe map registered sex offenders within the buffer (if you have the equivalent where you are) and all known hazards (I'd probably call unlit, unplowed or forest path hazardous), and then create maybe a heatmap of danger. All of this should be doable in something like QGIS. I now have this image of a thematic map in my head which features a frightened child walking through a freezing forest having to dodge cougars, wolves and pedobears. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
--- On Tue, 14/7/09, Ævar Arnfjörð Bjarmason ava...@gmail.com wrote: I now have this image of a thematic map in my head which features a frightened child walking through a freezing forest having to dodge cougars, wolves and pedobears. Sounds like a blockbuster when compared to the usual garbage in garbage out of hollywood :) ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
hi simon, i will seriously think about it. there's my program osmrender.pl which only needs an dijkstra algorithm and some more lines of code. then it would do. please look at the wiki for the map it produces. http://wiki.openstreetmap.org/wiki/Osmrender.pl of course this can be finetuned. ps: only thing i need is time... gerhard On Tue, 2009-07-14 at 13:20 -0400, si...@mungewell.org wrote: Has anyone done any work in the area of making maps representing 'distance to feature' as measure along routable roadways/footways? Local school board is trying to drop school bus service school kids (from grade 1 up) within 2.0km of school and get them to walk to school (bear in mind this is a rural community with -40'C winters, wildlife such as bears/cougars and an ungated CPR train route through it). I have some ideas about representing the data, but does anyone have any suggestions on how to parse through each node and measure the distance along each way to produce a list with 'distance to school' tag for every hightway node? Is this the 'best' approach? The node list would/could then be passed through a variant of 'strm2osm.pl' to produce a contour layer, and then merged/plotted with osmarender. I could also imagine that this 'work' might be usefull in 'distance to nearest [bus stop, post box, etc]' maps/datasets. Cheers, Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
another thought: it would also be possible to create additional nodes (where i.e. type=dist, dist=xyz) in xml format that could be fed into an existing osm file. the resulting file could be rendered by KOSMOS. all this would need is an additional rule for the new distance nodes. then you would have nicer maps :-)) gerhard On Tue, 2009-07-14 at 13:20 -0400, si...@mungewell.org wrote: Has anyone done any work in the area of making maps representing 'distance to feature' as measure along routable roadways/footways? Local school board is trying to drop school bus service school kids (from grade 1 up) within 2.0km of school and get them to walk to school (bear in mind this is a rural community with -40'C winters, wildlife such as bears/cougars and an ungated CPR train route through it). I have some ideas about representing the data, but does anyone have any suggestions on how to parse through each node and measure the distance along each way to produce a list with 'distance to school' tag for every hightway node? Is this the 'best' approach? The node list would/could then be passed through a variant of 'strm2osm.pl' to produce a contour layer, and then merged/plotted with osmarender. I could also imagine that this 'work' might be usefull in 'distance to nearest [bus stop, post box, etc]' maps/datasets. Cheers, Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
only needs an dijkstra algorithm and some more lines of code. sounds so simple... ;-) I had a look at the wikipedia page (http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm) and that would look to make sense. I suppose the way to do it would be to break the OSM highways down in unique ways which link junctions, so that each could be given a length (following node to node) and then use the dijkstra to calculate distance to each 'end' node. This would also be good for producing paper maps with distance values along the side of motoring maps. You just add the distances along chosen route. Not sure how this would work for multiple reference points, ie. nearest bus_stop. But that's not what I asked Thanks, Simon. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] 'Distance to feature' maps?
Good luck changing their minds. We have a similar rule near here - no bad winters or wild animals, but some kids were supposed to walk across a major multi-lane highway that had no crossing at all for vehicles for several km each way, but they were within a 1km circle of the school, so no transport. 2009/7/15 si...@mungewell.org: Local school board is trying to drop school bus service school kids (from grade 1 up) within 2.0km of school and get them to walk to school (bear in mind this is a rural community with -40'C winters, wildlife such as bears/cougars and an ungated CPR train route through it). ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
