Hi Frank,
Thanks for the reply. Your idea sounds good but i don't think it will work
for me. Really crazy stuff in that shapefile from OpenStreetMap, even
isolated polygons show up and many other things. I get it, its a raw file,
not a routable one but still...
I did find a solution that looks for the next 5 graph nodes that are closer
to the given destination and source points and then looks for a valid road
using as parameters combinations of those nodes. This option will fail if
the start point is somewhere near to a freaky polygon like this
http://screencast.com/t/oBA3aGpaqLy (which apparently is the interior alley
system of a residence), but lets hope the user of the app will not be so
crazy to give as the input data a point from a place like this.
On Tue, Sep 2, 2014 at 8:30 PM, Frank van der Hulst <drifter.fr...@gmail.com
> wrote:
> Hi Elena,
>
> I ran into this in my project too.
>
> My solution was to measure the distance from my tree to each of the
> candidate LineStrings. Only if the distance is less than a specified
> threshold would the new LineString be added. Depending on the quality of
> the source data, a threshold of between 20m and 150m worked for me. You
> need that distance anyway, because you need to find the nearest point of
> the tree to the new LineString, so it will be added in the right place. You
> could even do a first run through the data, deleting any LineString which
> doesn't connect to any other LineString. But that will be another O(n*n)
> process.
>
> One thing I did to improve speed (although I am only looking at a few
> hundred LineStrings at most) was to sort the list so that (usually) the
> best candidate would be found early. Criteria for the sort were things like
> the length of the candidate, the distance in the right direction, the
> distance from the nearest point on the tree to the destination.
>
> Frank
>
>
>
> On Tue, Sep 2, 2014 at 8:58 PM, Elena Pop <elena0...@gmail.com> wrote:
>
>> I am using a shapefile downloaded from OpenStreetMap in order to
>> calculate a route between two points. The graph created contains some
>> isolated arcs that are messing up my routing like these ones
>> http://screencast.com/t/4jnQFx57
>> Is there any way i can remove them in a fast way, before or after the
>> graph creation?
>>
>> The way i am implementing the project is creating a List of linestrings
>> read from the shapefile and then iterating through the list and adding each
>> linestring to the graph. I did find a solution but it is not an optimal one
>> since it takes more than 1 hour to get results.
>> My solution looks to count the number of occurrences of the first and he
>> last point of a linestring in the other linestrings and if one of the
>> nOccurrences is >1 the linestring will get added to the graph.
>>
>> Basically i will have to iterate twice through the list, but since the
>> list has more than 70000 linestrings it will do a O(n*n) program which will
>> take forever and i need to get results in ms not in hours.
>>
>> Is there other way that i can do this?
>>
>>
>> <http://screencast.com/t/4jnQFx57>
>>
>>
>> ------------------------------------------------------------------------------
>> Slashdot TV.
>> Video for Nerds. Stuff that matters.
>> http://tv.slashdot.org/
>> _______________________________________________
>> GeoTools-GT2-Users mailing list
>> GeoTools-GT2-Users@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users
>>
>>
>
------------------------------------------------------------------------------
Slashdot TV.
Video for Nerds. Stuff that matters.
http://tv.slashdot.org/
_______________________________________________
GeoTools-GT2-Users mailing list
GeoTools-GT2-Users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users
