Tweaks on point clouds

[cholmes]

Hey, maybe my understanding of point clouds is off, but this doc seems to talk about them as if they're points. As far as I know you wouldn't put point clouds in GeoJSON, or in geoparquet or flatgeobuf. Las and COPC are the format for point clouds (which are 3d points - https://en.wikipedia.org/wiki/Point_cloud 'A point cloud is a discrete set of data points in space. The points may represent a 3D shape or object. Each point position has its set of Cartesian coordinates (X, Y, Z)'

I think ideally the diagram would be tweaked too, to say 'point cloud' and maybe show things in 3d. In my view 'vector' is points, lines or polygons, so could maybe represent all 3.

Thanks for the work on this, other than that it looks great.

[kylebarron]

I agree it's a bit too muddled between "point (vector)" and "point (point cloud)" data. I've never heard of someone storing point cloud data in geojson, and I certainly would never want to recommend that.

As an aside, I started an updated doc for copc but never got around to finishing that yet

[kylebarron]

As far as I know you wouldn't put point clouds in ... geoparquet

Well.. I'm not an expert on point clouds, but pdal did implement support for geoparquet https://github.com/PDAL/PDAL/pull/4115 🤷‍♂️

[wildintellect]

I think we're all in agreement the line between when to use a file type for a data format is blurry. Example: sparse LiDAR from space (GEDI and ICESat2) could be stored in almost any file type (all the vector formats, COPC, and HDF5). Perhaps we need a better definition of what is a Point Cloud, and when you have enough data to warrant COPC or potentially Geoparquet. Most vector formats do allow XYZM coordinates (seen plenty of 3D data in Postgis too)...

How about putting an asterisk on GeoJSON (Not recommended for large data volumes)

[cholmes]

I think we're all in agreement the line between when to use a file type for a data format is blurry. Example: sparse LiDAR from space (GEDI and ICESat2) could be stored in almost any file type (all the vector formats, COPC, and HDF5). Perhaps we need a better definition of what is a Point Cloud, and when you have enough data to warrant COPC or potentially Geoparquet. Most vector formats do allow XYZM coordinates (seen plenty of 3D data in Postgis too)...

How about putting an asterisk on GeoJSON (Not recommended for large data volumes)

Yes, fully agreed that the line is blurry, and that you 'can' put many kinds of data in many kinds of formats - you can turn your raster grid to a series of points you put in GeoJSON. But I think this guide will be most useful if it gives a clear recommendation to people - you have X data use Y format. Could link to a more nuanced discussion, but to me the most valuable thing is to make it simple, don't put the complexity of choice on the user. So I would just position GeoJSON for points, lines and polygons, but not 'point clouds', as those typically are large data collections.

I am curious to hear more from Howard on GeoParquet for point clouds. But to me COPC is the format that is best geared towards it, the most tools will work with it, etc. Telling people to put their point cloud data into geoparquet will make it so they can't tap into a ton of tools to actually work with point clouds.

[kylebarron]

you can turn your raster grid to a series of points you put in GeoJSON. But I think this guide will be most useful if it gives a clear recommendation to people - you have X data use Y format

I agree 💯 . Even with that PDAL PR, I wouldn't suggest anyone to use GeoParquet for point data. I think it would be clearer if we totally removed a mention of geojson for point clouds

[zacharyDez]

Hey! What is the state of this PR? Can we go ahead and merge it? Changes seem ready to go. @cholmes @wildintellect @kylebarron