Category: Thoughts

Meshlab can do this, if you don’t have access to a linux machine and use

ctmconv red-rocks-smrf-only-delaunay.ply red-rocks-smrf-only-delaunay.ply

Meshlab seems great, may even be as useful as Blender.

On Windows 10, the default viewer.

I’m going to use the Creality Slicer (which is based off Cura). It doesn’t quite fit.

This seems reasonable to print.

Will fire up the printer and see!

Using Blender, I add a cube and played around with the positioning. I found the Blender UI to be very non-intuitive.

Then export as PLY file. It looks like this.

I was playing around with the lidar dataset from UCSD and visualizing it on Entwine.

Wanted to share some beautiful cool pictures. Lidar is very interesting. Here are the links to play this yourself.

https://usgs.entwine.io/data/view.html?r=%22https://s3-us-west-2.amazonaws.com/usgs-lidar-public/ARRA-CA_GoldenGate_2010%22

https://s3-us-west-2.amazonaws.com/usgs-lidar-public/ARRA-CA_GoldenGate_2010/ept.json

Looking at the data format, it the bounding boxes for the octo-tree here are:

Points

-13632855.894, 4553466.772, 1.042
-13635492.356, 4553196.558, 1.418
-13634390.511, 4555612.511, 72.595
-13633919.870, 4551994.285, 81.365

Ones in bold are the min/max vals, thus forming our X, Y bounds.

  {
      # type: readers.ept
      bounds: ([-13635492.356,-13632855.894], [4555612.511, 4551994.285])
      filename: https://s3-us-west-2.amazonaws.com/usgs-lidar-public/ARRA-CA_GoldenGate_2010/ept.json
      resolution: 0.5
      # up to 0.5m resolutions
    }

Actually this didn’t return any data. So I’m doing another selection with larger area.

The pipeline

{
  "pipeline": [
    {
      "bounds": "([-13637879,-13630062], [4549481,4557084])",
      "filename": "https://s3-us-west-2.amazonaws.com/usgs-lidar-public/ARRA-CA_GoldenGate_2010/ept.json",
      "type": "readers.ept"
    },
    {
      "filename": "golden-gate.laz",
      "type": "writers.las"
    }
  ]
}

474M May 4 23:50 golden-gate.laz

I put together and loaded this data, visualized with potree:

http://potree.entwine.io/data/custom.html?t=%5B-13633083.085886605,4558454.642095785,-3140.1667813895765%5D&p=%5B-13633789.913150359,4538864.068838491,6564.589746511545%5D&r=%5B%22http://127.0.0.1:8080/golden-gate.laz/ept.json%22,%22https://s3-us-west-2.amazonaws.com/usgs-lidar-public/ARRA-CA_GoldenGate_2010%22%5D&m=8&era=%5B-148,781%5D&cf=%5B0,3,4,5,6,7,8,12%5D

That’s it folks!

Source Data

Looking at Red Rocks.

PDAL pipeline

# This is a hjson file, https://hjson.org/
# Linux bash
#GET=https://github.com/hjson/hjson-go/releases/download/v3.0.0/linux_amd64.tar.gz
# macOS bash
#GET=https://github.com/hjson/hjson-go/releases/download/v3.0.0/darwin_amd64.tar.gz
# Install
#curl -sSL $GET | sudo tar -xz -C /usr/local/bin
# Translate to Json
#hjson -j pipeline.hjson > pipeline.json
#pdal pipeline pipeline.json --verbose 8

{
  pipeline:
  [
# Input
    {
      # read from our ept server
      # up to 0.5m resolutions
      # type: readers.ept
      # bounds: ([802000, 802500], [2493000, 2493500])
      # filename: http://localhost:8080/ept.json
      filename: red-rocks.laz
      # filename: http://na.entwine.io/red-rocks/ept.json
      # resolution: 0.5
    }
#     {
#       # read from our las file
#       type: readers.las
#       filename: small500-no-outliers.laz
#     }


# Filters
    {
      # adds a classification value of 7 to the noise points
      type: filters.outlier
      # method: radius
      # radius: 1.0
      # min_k: 8 # min number of neighbors in radius

      method: statistical
      mean_k: 8
      multiplier: 3

    }

    {
      # voxel-based sampling filter
      # reduce the size of the pc
      # cell size of 0.2 meters in xyz
      type: filters.voxelcenternearestneighbor
      cell: 0.1
    }

    {
      # Need to assign point cloud dimension NumberOfReturns 1
      # Otherwise: "No returns to process."
        type:filters.assign
        assignment : NumberOfReturns[0:0]=1
    }

    {
      # Ground classification, ignore the noise points
      type: filters.smrf
      ignore:Classification[7:7]
    }

    {
      # only allow ground classified points
      type: filters.range
      limits: Classification[2:2]
    }

    {
      # OPTIONAL
      # turn this into a DEM 3D model
      # do not use multiple types
      # type: filters.delaunay
      type: filters.poisson
    }


# Output

# # OPTIONAL PLY IF DEM
    {
      # write to ply
      type:writers.ply
      filename: red-rocks-smrf-only-poisson.ply
      faces:true
      storage_mode: default
    }
# Output
    # {
    #   # write to laz
    #   type:writers.las
    #   filename: red-rocks-ground.laz
    # }
  ]
}

https://gist.github.com/sunapi386/9a9ece302d646ee80a72fc494423a633

Mesh Results

Greedy Projection

Poisson

To go a little more “detailed”, I put depth to 12. The file went from 84M to 789M. And it is definitely overkill for 3D printing.

Grid Projection

Program failed to compute the grid projection.

 pdal pipeline dtm-gdal.json --verbose 8 
(PDAL Debug) Debugging...
(pdal pipeline Debug) Attempting to load plugin '/usr/local/lib/libpdal_plugin_filter_gridprojection.so'.
(pdal pipeline Debug) Loaded plugin '/usr/local/lib/libpdal_plugin_filter_gridprojection.so'.
(pdal pipeline Debug) Initialized plugin '/usr/local/lib/libpdal_plugin_filter_gridprojection.so'.
(pdal pipeline readers.las Debug) GDAL debug: OGRSpatialReference::Validate: No root pointer.
(pdal pipeline readers.las Debug) GDAL debug: OGRSpatialReference::Validate: No root pointer.
(pdal pipeline readers.las Debug) GDAL debug: OGRSpatialReference::Validate: No root pointer.
(pdal pipeline Debug) Executing pipeline in standard mode.
(pdal pipeline filters.gridprojection Debug) 		Process GridProjectionFilter...
[pcl::GridProjection::getBoundingBox] Size of Bounding Box is [5.500000, 6.000000, 5.000000]
[pcl::GridProjection::getBoundingBox] Lower left point is [-2.500000, -2.500000, -2.500000]
[pcl::GridProjection::getBoundingBox] Upper left point is [3.000000, 3.500000, 2.500000]
[pcl::GridProjection::getBoundingBox] Padding size: 3
[pcl::GridProjection::getBoundingBox] Leaf size: 0.500000
(pdal pipeline filters.gridprojection Debug) 		3141373 before, 180 after
(pdal pipeline filters.gridprojection Debug) 		180
double free or corruption (!prev)
fish: “pdal pipeline dtm-gdal.json --v…” terminated by signal SIGABRT (Abort)

Cura 3D Print Slice

Cura can take STL inputs. Converting the PLY into STL is simple.

sudo apt install openctm-tools

Then ctmconv red-rocks-smrf-only-delaunay.ply red-rocks-smrf-only-delaunay.stl can convert ply to stl

ctmviewer red-rocks-smrf-only-delaunay.ply visualizes the ply. Which is what I used above.

Looks like the Poisson is prettier.

I’ll continue writing this latter, until I have something printed. 🙂

https://pdal.io/stages/filters.voxelcenternearestneighbor.html#filters-voxelcenternearestneighbor

The VoxelCenterNearestNeighbor filter is a voxel-based sampling filter. The input point cloud is divided into 3D voxels at the given cell size. For each populated voxel, the coordinates of the voxel center are used as the query point in a 3D nearest neighbor search. The nearest neighbor is then added to the output point cloud, along with any existing dimensions.

Notice the red dots are much more sparse than the gray intensity dots. Red dots are separated 1.0 meters and gray are 0.1 meters.

To generate this I did converted with PDAL and then used Potree to visualize.

pdal pipeline $HOME/voxcnn-1.0.json 

voxcnn-1.0.json
 [
     "801403-802580-2493384-2494335.laz",
    {
         "type":"filters.voxelcenternearestneighbor",
         "cell":1.0
     },
     "801403-802580-2493384-2494335-voxelcenternearestneighbor-1.0.laz"
 ]

-rw-rw-r-- 1 jsun jsun  65M Apr 11 21:01 801403-802580-2493384-2494335-voxelcenternearestneighbor.laz
 -rw-rw-r-- 1 jsun jsun 8.1M Apr 11 21:04 801403-802580-2493384-2494335-voxelcenternearestneighbor-1.0.laz