Run PDAL in a Docker Container...

This is a tutorial to demonstrate how to containerize and run PDAL. PDAL is a stand-alone software package that can analyze and manipulate point cloud data files such as .las and .laz. In this tutorial, we will filter a Cloud Optimized point cloud (.copc.laz) to classify the ground points from the tree canopy points. The point cloud file is located in this repository and is called odm_georeferenced_model.copc.laz. It currently has no point classification. We will be using a Cloth Simulation Filter. Background on Coth Simulation Filter can be found here

1. Clone this repository to your local machine

git clone https://github.com/jeffgillan/pdal_filters.csf.git

2. Change directories into the newly cloned repository

cd pdal_filters.csf

3. Run the Container

docker run -v $(pwd):/data jeffgillan/pdal_csf:1.0

Your if everything worked correctly, you should have a new file odm_georeferenced_model_filtered.copc.laz in your present working directory.

You can visualize and see the classification using https://viewer.copc.io/

---

How to Build this Docker Container

1. PDAL Pipeline

Analyzing pointclouds in PDAL requires users to specify processing steps within a json file. PDAL uses the term 'pipeline' to describe this json file.

Create an empty json file

touch filter_csf.json

Open the pipeline file

nano filter_csf.json

Write this info in the pipeline file. It will classify the points as ground or non-ground. It will output as a cloud optimized point cloud (.copc.laz).

[
    "file.las",
    {
        "type":"filters.csf"
    },
    "output.copc.laz"
]

2. Create a shell script

Because PDAL is a command line tool, we need to instruct it with shell commands. The shell script will loop through a directory (within the container) and find all .copc.laz. It will pass the files to the json pipeline which will execute the classification of ground v. non-ground points. The path to the json needs to be specified within the shell script.

touch pdal_csf.sh

Edit the shell script

nano pdal_csf.sh

#!/bin/bash

# Define the pipeline JSON file
pipeline="/app/filter_csf.json"

# Loop over LAS/LAZ files in a directory and subdirectories. It this example, it loops over `/data` in the container. 
find /data -type f \( -name "*.copc.laz" \) -print0 | while IFS= read -r -d '' file; do
    # Get the file extension
    extension="${file##*.}"

    # Run the pipeline with the appropriate reader based on the file extension. Output the file as `<file_name>_filtered.copc.laz`

    pdal pipeline -i "$pipeline" --readers.las.filename="$file" --writers.copc.filename="${file%.copc.laz}_filtered.copc.laz"
done 

3. Create a Dockerfile

You are creating a Docker image that includes a PDAL base image, the json pipeline file, and the shell script.

Create an empty Dockerfile

touch Dockerfile

Edit the Dockerfile

nano Dockerfile

FROM pdal/pdal:sha-597ab2df

WORKDIR /app

COPY pdal_csf.sh /app/pdal_csf.sh

COPY filter_csf.json /app/filter_csf.json

RUN chmod +x pdal_csf.sh

ENTRYPOINT ["/app/pdal_csf.sh"]

The following is happening in the Dockerfile:

A PDAL base image is being pulled from Dockerhub

I set the working directory of the container to /app

I copy in the shell script to the path /app

I copy in the pipeline json file to the path /app

I run chmod +x on the shell script to give everyone permissions

The entrypoint is where the container starts. I want it to start with the shell script.

4. Build the docker image

You are telling it to build an image with the name 'jeffgillan/pdal_csf' with the tag '1.0'. You are building from the Dockerfile in the current working directory '.'

docker build -t jeffgillan/pdal_csf:1.0 .

5. Run the container

You are mounting a local volume (-v) directory to the container (/data). This local directory should have all of the point clouds files you want to convert. $(pwd) is telling it that the point clouds are in the current working directory. Alternatively, you could specify the point clouds are locating in any local directory.

docker run -v $(pwd):/data jeffgillan/pdal_csf:1.0

6. Outputs

The tool should output <file_name>_filtered.copc.laz files to the same directory where the input point clouds were stored.

7. Upload Image to Docker Hub

docker push jeffgillan/pdal_csf:1.0

____