[QGIS-Developer] Point cloud renderer: tile borders visible?

[Nyall Dawson]

On Tue, 8 Dec 2020 at 23:59, Andreas Neumann <andreas at qgis.org> wrote:
>
> Hi Martin,
>
> Thanks for the explanation.
>
> I confirm that reducing the maximum error helps to remove the artefacts. WIth 0.5 mm I could still see them at one particular zoom level, but with 0.4 mm it was completely gone.

In https://github.com/qgis/QGIS/pull/40504 I've changed the default to 0.3.

Thanks for the feedback!
Nyall


>
> Thanks a lot,
> Andreas
>
>
> On Tue, 8 Dec 2020 at 13:15, Martin Dobias <wonder.sk at gmail.com> wrote:
>>
>> Hi Andreas
>>
>> The artifacts you are seeing are caused by the fact how the Entwine tool works in the default configuration. When creating the octree index of point cloud data, points are organized in a hierarchy of nodes (each containing usually tens of thousands of points), with each level of the hierarchy having smaller bounding volume and higher density. One optimization that Entwine does is that if child nodes contain a small amount of points, they get merged into the parent node in the hierarchy - in the hope this will lower the number of nodes and therefore also the number of requests that need to be done. See Entwine configuration [1] for more details - especially the options like minNodeSize, maxNodeSize, overflowDepth, overflowThreshold. So the artifacts could be simply fixed by reindexing the dataset with modified configuration.
>>
>> All that said, maybe none of that is necessary:
>> 1. I would suggest lowering "Maximum error" in Layer styling to e.g. 0.5 mm or even 0.3 mm (assuming the default point size of 1mm) - this should be giving you a rendering result with very little holes - it is much nicer to look at and the artifacts should be gone at all scales  (it will increase the rendering time, but I think it is worth it and I think we should lower the default maximum error - the higher maximum error was initially used mainly as a debugging tool for us to see if the point spacing in the rendered map matches the expectations)
>> 2. For indexing of point clouds, we will be internally using Untwine [2] - a newer approach (also coming from friends at Hobu!) that has different configuration and probably won't be showing these artifacts. Peter is currently working on the integration [3]
>>
>> Regards
>> Martin
>>
>> [1] https://entwine.io/configuration.html
>> [2] https://github.com/hobu/untwine
>> [3] https://github.com/qgis/QGIS/pull/40404
>>
>>
>> On Tue, Dec 8, 2020 at 10:30 AM Andreas Neumann <a.neumann at carto.net> wrote:
>>>
>>> Hi,
>>>
>>> When testing the 2D point cloud renderer I noticed that in certain scales / zoom levels one can clearly see point cloud tile borders:
>>>
>>> see screenshot at https://www.carto.net/neumann/temp/point_cloud_artefacts.png
>>>
>>> There seem to be different densities involved in certain regions here, but clearly they correspond to tile borders.
>>>
>>> Is this an expected artefact? Can this be avoided?
>>>
>>> When I zoom in or out, these artefacts seem to disappear.
>>>
>>> Thanks and greetings,
>>>
>>> Andreas
>>>
>>>
>>> _______________________________________________
>>> QGIS-Developer mailing list
>>> QGIS-Developer at lists.osgeo.org
>>> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>>> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>>
>> _______________________________________________
>> QGIS-Developer mailing list
>> QGIS-Developer at lists.osgeo.org
>> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>
>
>
> --
>
> --
> Andreas Neumann
> QGIS.ORG board member (treasurer)
> _______________________________________________
> QGIS-Developer mailing list
> QGIS-Developer at lists.osgeo.org
> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer