# Geometry :

Subject: Re: [geometry] intersection for two segments in 3D failed
Date: 2015-03-24 09:59:12

Hi,

Hmfkz wrote:
> Hi,
>
> I tried to use boost geometry for intersection of the two line segments in
> 3D.
> Here

> is the link of question in stackoverflow.
>
> I am wondering if it is possible to do intersection for segments in 3D. if
>
> I am using boost version 1.53.
>

Sorry, 3D intersection is not yet properly supported. I think it works
properly only with segments containing points with z coordinate equal to 0.

But yes, in the case of Linear/Linear geometries combinations (e.g.
Segment/Segment, Linestring/Linestring, etc.) it should be possible to
calculate the intersection in pure 3D cartesian. Linear/Areal (e.g.
Linestring/Polygon) would probably be more complicated since for each
Polygon we'd need to calculate sides (that's an implementation detail)
WRT the Polygon's surface.

Could you please fill a ticket with the feature request here
https://svn.boost.org/trac/boost/newticket so we won't forget about it?

But do you really need to find an intersection of segments in 3D? Are
those segments arbitrary distributed in 3D or are they e.g. lying on
some plane? If the latter was true, you could first project those
segments into the plane and then find the intersection of 2d segments.

In the case of "true" 3d cartesian segment/segment you could try to
implement it by yourself.
E.g. see: http://www.realtimerendering.com/intersections.html (ray/ray
case).
Here is a formula: http://www.realtimerendering.com/intersections.html#I304
where
o1 = segment1.p1
d1 = normalized(segment1.p2 - segment1.p1)
o2 = segment2.p1
d2 = normalized(segment2.p2 - segment2.p1)
(o states for ray's "origin", d for "direction")
And there are some special cases which must be handled like parallel
rays or closest points outside a segment (t < 0 || t > length(segment)).

Regards,