[pdal] Delete low points below surface

[Howard Butler]

It would be slow to compute, but one thought is if you had an additional characterization from filters.covariancefeatures https://pdal.io/en/stable/stages/filters.covariancefeatures.html you could craft a filters.expression that that uses those attributes to help you identify the "linear tunnels".

> On Nov 10, 2025, at 7:54 AM, Roland Spielhofer via pdal <pdal at lists.osgeo.org> wrote:
> 
> Hi,
> I tried cell = 0.02 - 0.04 and thresholds of 0.001 to 0.01.
> In the second image it is quite good to see that the sinusoidal line of points is quite dense in z-direction and I guess that is the problem that the threshold does not work well. It must be quite small, but larger than the variance in the surface.
>  
> In the meantime I also tried this:
>  
> {
>       "type": "filters.gridDecimation",
>       "output_type":"max",
>       "resolution": "0.03",
>       "value":"Classification=9"
>   },
>  
> Keep only the highest of a grid cell - with resolution of 0.03 the low points go away, with 0.025 or 0.02 they stay - 0.02 is roughly the distance between the transversal scan lines. Hoewever, it decimates the concrete surface as well by a mangitude, which I would try to avoid.
>  
> Regards,
> Roland
> Gesendet: Montag, 10. November 2025 um 14:37
> Von: "Bradley Chambers" <brad.chambers at gmail.com>
> An: "Roland Spielhofer" <rspiel at gmx.net>
> CC: pdal at lists.osgeo.org
> Betreff: Re: [pdal] Delete low points below surface
> I’d be interested to know what ELM parameters you used. It is configurable, though perhaps still difficult to fine tune. 
> 
> On Mon, Nov 10, 2025 at 06:38 Roland Spielhofer via pdal <pdal at lists.osgeo.org <mailto:pdal at lists.osgeo.org>> wrote:
>> Hi,
>> I am loking for a PDAL pipline to get rid of low lying points.
>> The point cloud I am working with is very dense and represents the surface of a concrete pavement.
>> The slabs of the pavement are divided by joints - and the points lying in these joints just below the surface I want to get rid of.
>>  
>> I created some images showing details of one joint:
>> https://imgur.com/a/oX7dZcE
>>  
>> The surface is very flat, and the point density is ~ 5 mm in transversal and 20 mm in longitudinal direction (=driving direction).
>> The joints have a depth of ~40 mm (see images).
>>  
>> I tried out the elm filter, but I guess that the relative z-difference between the points is too little to mark them as noise.
>>  
>> Any idea to filter the points in the joint with PDAL would be appreciated (I can also provide the LAZ pointcloud itself for experiments).
>>  
>> Thanks in advance and best regards,
>> Roland
>> _______________________________________________
>> pdal mailing list
>> pdal at lists.osgeo.org <mailto:pdal at lists.osgeo.org>
>> https://lists.osgeo.org/mailman/listinfo/pdal
> _______________________________________________
> pdal mailing list
> pdal at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/pdal

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/pdal/attachments/20251111/5e9180ef/attachment-0001.htm>