[R-sig-Geo] Advice on clustering LiDAR point clouds

[Wesley Roberts]

Dylan & Etienne,

Many thanks for your suggestions, I am looking for tree location as part of a tree counting analysis in a plantation forest. The espacement of the trees is 2,5 m (within rows) by 3 m (between rows) so the plantation is fairly dense. clus from the spatclus package did not turn out to be a useful approach for my application as it only ever found one cluster. I thought of iteratively feeding it the single tree tops with the population data (all points) but it didnt seem to like that and would not run with only the one data point. In the end I have used ArcGIS to create TINS and then rasters which I will analyse in eCognition. Applications in R seem very specific and I don't have the experience in R to adapt them to my application, still I like R and will use it where I can in the future.

Regards,
Wesley

>>> Dylan Beaudette <debeaudette at ucdavis.edu> 18/08/2009 21:33 >>>
On Tuesday 18 August 2009, Etienne Bellemare Racine wrote:
> Hi Wesley,
>
> It is unclear to me if you want to aggregate points to distinguish
> individual tree, or if you want to locate the treetops ?
>
> Anyway, I suggest you first use your grid to locate treetops (using
> maximum value as you said), then clustering using the treetops. Is your
> plantation dense ? because if the crowns are touching each other, I
> think it will be pretty hard to cluster them using only points. If you
> have treetops, you can aggregate points around individuals treetops.
>
> To get maximum values, maybe sp::overlay can get you started, but I
> think it is more to get grid value to point than reverse. I do it in
> ArcGIS, but I would be happy to have a solution in R if you find one.
>
> Etienne

Also see r.in.xyz within GRASS GIS.

Dylan

>
> Wesley Roberts a crit :
> > Dear R-sig-Geo,
> >
> > I am currently looking at clustering a LiDAR point cloud (trees in a
> > plantation forest) using R and have some questions that I hope some of
> > you may be able to answer.
> >
> > My method is a two stage approach, firstly I selected potential tree
> > locations by overlaying a static grid on the point data and selecting the
> > maximum value within each grid. These locations were stored as potential
> > tree locations and have been used as sample data in a spatial clustering
> > approach (I would have liked to use a moving filter but could not find a
> > local maximum approach implemented in R). These points and the original
> > data are now being clustered using the clus algorithm in the spatclus
> > package.
> >
> > My query regards the use of an algorithm developed for disease mapping
> > (Kuldorff's circular zone in 2D) with Lidar data. The density of the
> > lidar points is around 5 per square meter and I am concerned that the
> > algorithm will not be able to identify clusters based on height. I am yet
> > to inspect the results as the clus algorithm is still running so I cant
> > comment on that right now, but I was wondering if anyone on the list had
> > any suggestions wrt the clustering and or segmentation of lidar point
> > clouds using R. I am unwilling to use interpolation as I want to avoid
> > the lengthy process of selecting the correct interpolation procedure and
> > or model and would like to stick with the point cloud.
> >
> > Any advice on this matter would be greatly appreciated.
> > Many thanks and kind regards,
> >
> > Wesley
> >
> >
> > Wesley Roberts MSc.
> > Researcher: Earth Observation
> > Natural Resources & the Environment (NRE)
> > CSIR
> > Tel: +27 (0)21 888-2490
> > Fax: +27 (0)21 888-2693
> > "To know the road ahead, ask those coming back."
> > - Chinese proverb
>
> 	[[alternative HTML version deleted]]



-- 
Dylan Beaudette
Soil Resource Laboratory
http://casoilresource.lawr.ucdavis.edu/ 
University of California at Davis
530.754.7341



-- 
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard. 
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.

This message has been scanned for viruses and dangerous content by MailScanner, 
and is believed to be clean.  MailScanner thanks Transtec Computers for their support.