Filtering data with PCL

Introduction

PDAL is both a C++ library and a collection of command-line utilities for data processing operations. While the PDAL library addresses point cloud exploitation and filtering, this takes a back seat to its primary objective of being a data translation library, helping developers to navigate the a wide variety of point cloud formats. PCL is another C++ library that is focused on developing a rich set of point cloud processing routines, with less of a focus on formats and data translation. Acknowledging this, the PCL Block filter was developed to serve as a bridge between the two libraries, enabling rapid development of point cloud processing pipelines.

See also

See filters.pclblock for details on PDAL’s PCL Block filter.

Quick Start

The Quickstart document describes how to use PDAL with Docker, which includes built-in PCL support. After you have worked through that document, you should be able to run any PDAL PCL operations.

PDAL Pipeline kernel

Note

A full description of the PDAL pipeline concept is beyond the scope of this tutorial but the Pipeline, pipeline, and Reading with PDAL documents contain detailed examples and background information.

The filters.pclblock is implemented as a PDAL filter stage and as such is easily accessed via the PDAL pipeline. It accepts a single, required option - the name of the JSON file describing the PCL Block.

A sample pipeline JSON which reads/writes LAS and has a single PCL Block filter is shown below.

{
  "pipeline":[
    "autzen-point-format-3.las",
    {
      "type":"filters.pclblock",
      "filename":"passthrough.json"
    },
    "foo.las"
  ]
}

And is run from the command line thusly.

$ pdal pipeline passthrough.json

This simple pipeline reads the input LAS (autzen-point-format-3.las), passes it through the PCL Block (passthrough.json), and writes the output LAS (foo.las).

When run, it should produce output similar to this:

Requested to read 106 points
Requested to write 106 points
0
Processing /home/vagrant/pdal/test/data/filters/pcl/passthrough.json

--------------------------------------------------------------------------------
NAME:   PassThroughExample ()
HELP:
AUTHOR:
--------------------------------------------------------------------------------
106 points copied

   Step 1) PassThrough

      Field name: z
      Limits: 410.000000, 440.000000

76(writers.las DEBUG: 3): Wrote 81 points to the LAS file
.100

PDAL PCL kernel

For users that would like to bypass the creation (and subsequent modification) of the pipeline JSON for every file they wish to process, there is another option: the pdal pcl command.

$ pdal pcl -i /path/to/input/las -p /path/to/pcl/block/json -o /path/to/output/las

This is functionally equivalent to the original pdal pipeline command, but does not afford the flexibility of constructing the pipeline (i.e., none the other PDAL filters are accessible).

The same can be accomplished with the pdal pcl command. The basic syntax for the command is

$ pdal pcl -i <input cloud> -p <PCL Block JSON> -o <output cloud>

where the JSON file specified with -p is the same file that would be embedded in the pipeline JSON file. This can be useful when the pipeline does not change frequently, but the input/output filenames do.

For example, the above pdal pipeline example can be written with pdal pcl like this:

$ cd pdal  # your PDAL source tree
$ cd test/data
$ ../../bin/pdal pcl -i autzen/autzen-point-format-3.las -p filters/pcl/example_PassThrough_1.json -o ../temp/foo.las -v4

This should produce the output

Requested to read 106 points
Requested to write 106 points
0
Processing /home/vagrant/pdal/test/data/filters/pcl/passthrough.json

--------------------------------------------------------------------------------
NAME:   PassThroughExample ()
HELP:
AUTHOR:
--------------------------------------------------------------------------------
106 points copied

   Step 1) PassThrough

      Field name: z
      Limits: 410.000000, 440.000000

76(writers.las DEBUG: 3): Wrote 81 points to the LAS file
.100

Examples

Simple point cloud cropping

The power of the PCL Block is really exposed through the JSON description. In this example, we apply a single PCL filter to the PointView. The PassThrough filter removes points that lie outside a given range for the specified dimension. Here, we are asking PCL to crop the input point cloud, returning only those points with z values in the range 100 to 200.

[
    {
        "name": "PassThrough",
        "setFilterFieldName": "z",
        "setFilterLimits":
        {
            "min": 410.0,
            "max": 440.0
        }
    }
]

(This example is taken from the unit test PCLBlockFilterTest_example_PassThrough_1.)

Point cloud cropping with outlier removal

Building on the previous example, we can string together multiple PCL filtering stages, such as the StatisticalOutlierRemoval filter. Note that the name field identifies the PCL filter by its class name, and furthermore that as of now only a handful of the PCL filtering options are accessible through the PCL Block. Similarly, select parameters of these classes can be set by specifying their public member functions by name.

[
    {
        "name": "PassThrough",
        "help": "filter z values to the range [410,440]",
        "setFilterFieldName": "z",
        "setFilterLimits":
        {
            "min": 410.0,
            "max": 440.0
        }
    },
    {
        "name": "StatisticalOutlierRemoval",
        "help": "apply outlier removal",
        "setMeanK": 8,
        "setStddevMulThresh": 0.2
    }
]

(This example is taken from the unit test PCLBlockFilterTest_example_PassThrough_2.)

Ground return filtering

The Progressive Morphological Filter (PMF) is an openly published approach to identifying ground vs. non-ground returns in point cloud data. An implementation of PMF is included with PCL and accessible through the PDAL’s PCL Block filter.

A complete description of the algorithm can be found in the article “A Progressive Morphological Filter for Removing Nonground Measurements from Airborne LIDAR Data” by K. Zhang, S. Chen, D. Whitman, M. Shyu, J. Yan, and C. Zhang.

To run the PMF with default settings, the PCL Block JSON is simply:

[
    {
        "name": "ProgressiveMorphologicalFilter"
        "setMaxWindowSize": 200,
    }
]

Additional parameters can be set by advanced users:

[
    {
        "name": "ProgressiveMorphologicalFilter",
        "setCellSize": 1.0,
        "setMaxWindowSize": 200,
        "setSlope": 1.0,
        "setInitialDistance": 0.5,
        "setMaxDistance": 3.0,
        "setExponential": true
    }
]

(These examples are taken from the unit tests PCLBlockFilterTest_example_PMF_1 and PCLBlockFilterTest_example_PMF_2.)

See here for a more detailed explanation of the PMF parameters.