[postgis-users] How to match lines between two similar graphs?

[Nicolas Ribot]

> Hi All,
>
> I have an interesting problem I am trying to solve and would love some
> feedback on how to best go about it.
>
> I have road data from two different vendor data sets. But this could also be
> one road network and a GPS track, so I think this is a pretty common use
> case.
>
> Assumptions:
>
> o the networks are similar, ie: they have similar roadway coverage
> o the two sets might be slightly misaligned, ie: shifted by some amount
> o the segments in the two data sets do not have to be broken into equivalent
> segments, ie: one segment in A might be represented my multiple segments in
> B
> o segments are not aligned end point wise, ie: a segment in A might go from
> mid-point one segment in B to the midpoint of a connected segment in B
> o in many cases I will be working with a set of lines in one set that I need
> to match to the other to select a matching set of lines.
>
> So strategies for matching these:
>
> 1. take a segment from A and buffer it, then intersect the data in B and
> select the longest intersected object. I can probably throw out any pieces
> smaller than the buffer distance.
>
> 2. Do the same but buffer and union the set of lines into a single
> multipolygon, and intersect that with the other set.
>
> 3. ??? Other ideas?
>
> Thoughts on performance?
>
> Typically I will have a small set (1-20) of segments to compare against a
> larger (100K-2M) set. Obvious a spatial index will will be used. But I'm
> wondering what is the fast way to do this matching computationally. I think
> I will want to be able to compare 1-200 sets like this every 5 mins as data
> comes in from a feed, while supporting other queries.
>

With this small input set, maybe st_hausdorffDistance() could be
useful to select good candidates in the larger set.

Nicolas