[postgis-users] Distance between two furthest points of a group

Mon Jul 13 08:41:33 PDT 2015

Thanks for the replies, interesting stuff.

I ended up using ST_ConvexHull(st_collect(geom)) in PostGIS (nice and fast) to create a polygon covering all the points, and then I used FME to measure the distance between all the vertices of that hull and the greatest distance was my answer. The hull being a much smaller dataset, it was much faster than it would have been had I tried to measure the billions of potential distances between millions of points.

I suspect the later part can be done in PostGIS too, but I needed to do some other bits in FME with the data, and as I know it much better than PostGIS, that seemed like the expedient option, ST_LongestLine in particular from the responses I saw here.

Thanks,
Jonathan

From: postgis-users-bounces at lists.osgeo.org [mailto:postgis-users-bounces at lists.osgeo.org] On Behalf Of Rémi Cura
Sent: Tuesday, June 30, 2015 8:11 AM
To: PostGIS Users Discussion
Subject: Re: [postgis-users] Distance between two furthest points of a group

Maybe you can even reduce this with cte :

with collected_geom AS (
select st_collect(geom) as geoms
from your_points
)
SELECT ST_LonguestLine(t1.geoms,t2.geoms)
FROM collected_geom AS t1, collected_geom AS t2
Cheers,
Rémi-C

2015-06-30 8:48 GMT+02:00 Nick Ves <vesnikos at gmail.com<mailto:vesnikos at gmail.com>>:
Didn't know about st_LongestLine.

Just tried and it amazed me!

on a dataset of 220k points (on a projected crs) it took ~ 1.5 secs to answer the querry :

select 1 as id,  ST_LongestLine(st_collect(geom),st_collect(geom)) geom from points ;

N

On Sat, Jun 27, 2015 at 2:41 PM, Nicklas Avén <nicklas.aven at jordogskog.no<mailto:nicklas.aven at jordogskog.no>> wrote:

Hallo

I haven't followed the whole conversation.

But one way is to collect alk points and usr ST_maxdistance on the colnections. or st_longestline. longestline will return a line where the end points is the two points furthest from eath other. Those functions is quite fast.

/Nicklas

Sent from my Cat® phone.
Den 27 jun 2015 13:25 skrev Rémi Cura <remi.cura at gmail.com<mailto:remi.cura at gmail.com>>:
Maybe I'm wrong, but your 2 farthest points should be on the boundary of the maximum bounding circle (feels right but couldn't prove it).
Thus you would compute this circle, then filter points not too far from it, then take the points with the max distance using an inner join (same as Nick, but you can save half the computation because dist(A,B)=dist(B,A), so simply add a condition a.id<http://a.id><b.id<http://b.id>)).
Another solution is to use bbox n nearest neighbour, which is indexed.
You wouldn't car too much about using bbox, because for points it only reduce precision to float instead of double.
This would be like :

SELECT a.id<http://a.id>,b.id<http://b.id>, st_distance(a.geom,b.geom) AS d
FROM my_points AS a , my_points AS b
ORDER BY a.geom <-> b.geom DESC
LIMIT 1
This is the probably the better easiness/speed ratio.

Cheers,
Rémi-C

2015-06-26 22:06 GMT+02:00 Nick Ves <vesnikos at gmail.com<mailto:vesnikos at gmail.com>>:
You can cross join to create the cartesian product of them and use it to calculate the distance of each with regards to the other:

select a.id<http://a.id>,b.id<http://b.id>, st_distance(a.geom,b.geom) d from points a cross join points b order by d desc limit 1;

ofc that will take forever because it will have to create an m x n table (800 secs and counting...)

As I see it the two points with the furthest distance between them should touch the borders of you datasets convexhull. So you can filter out those inside the boundaries and do the calculations with  the remaining points along the borders :

with f as
(
select a.geom,a.id<http://a.id> from
points foo,
(select ST_ExteriorRing (st_convexhull(st_collect(geom))) geom from points) bar
where st_Dwithin(foo.geom,bar.geom,0.00000001)
)
select a.id<http://a.id>,b.id<http://b.id>, st_distance(a.geom,b.geom) d from f  a cross join f b order by d desc limit 1;

should give you the id of your targets and the distance between them

On Fri, Jun 26, 2015 at 2:54 PM, Jonathan Moules <J.Moules at hrwallingford.com<mailto:J.Moules at hrwallingford.com>> wrote:
Hi List,
I have sets of points (up to 250,000 in a set) and I want to get the furthest distance between any of them.

In theory the simplest way is to use ST_MinimumBoundingCircle(ST_Collect(geography) and then get the diameter of that.

The problem is – I don’t seem to be able to get the diameter of that circle (which would give me the distance I want).

Does anyone have any thoughts on this? Is there a good way to get the diameter? Or some other way of getting the distance I desire.

Thanks,
Jonathan

________________________________

HR Wallingford and its subsidiaries uses faxes and emails for confidential and legally privileged business communications. They do not of themselves create legal commitments. Disclosure to parties other than addressees requires our specific consent. We are not liable for unauthorised disclosures nor reliance upon them.
If you have received this message in error please advise us immediately and destroy all copies of it.

HR Wallingford Limited
Howbery Park, Wallingford, Oxfordshire, OX10 8BA, United Kingdom
Registered in England No. 02562099

________________________________

_______________________________________________
postgis-users mailing list
postgis-users at lists.osgeo.org<mailto:postgis-users at lists.osgeo.org>
http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-users

_______________________________________________
postgis-users mailing list
postgis-users at lists.osgeo.org<mailto:postgis-users at lists.osgeo.org>
http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-users

_______________________________________________
postgis-users mailing list
postgis-users at lists.osgeo.org<mailto:postgis-users at lists.osgeo.org>
http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-users

_______________________________________________
postgis-users mailing list
postgis-users at lists.osgeo.org<mailto:postgis-users at lists.osgeo.org>
http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-users

This message has been scanned for viruses by MailControl<http://www.mailcontrol.com/>, a service from BlackSpider Technology

Click here<https://www.mailcontrol.com/sr/8wU+ANsngJbGX2PQPOmvUtzroFmuoUEujA1tmRx6QBJ9hPG3sWmR7sOI52Ay98QASuwfZeMf7FcAXoNDtAu6BA==> to report this email as spam.

________________________________

HR Wallingford and its subsidiaries uses faxes and emails for confidential and legally privileged business communications. They do not of themselves create legal commitments. Disclosure to parties other than addressees requires our specific consent. We are not liable for unauthorised disclosures nor reliance upon them.
If you have received this message in error please advise us immediately and destroy all copies of it.

HR Wallingford Limited
Howbery Park, Wallingford, Oxfordshire, OX10 8BA, United Kingdom
Registered in England No. 02562099

________________________________
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20150713/2bd2a358/attachment.html>