PDAL

PDAL Python support allows you to process data with PDAL into Numpy arrays. It provides a PDAL extension module to control Python interaction with PDAL. Additionally, you can use it to fetch schema and metadata from PDAL operations.

Installation

Note The PDAL Python bindings require the PDAL base library installed. Source code can be found at https://pdal.io and GitHub.

PyPI

PDAL Python support is installable via PyPI:

pip install PDAL

Developers can control many settings including debug builds and where the libraries are installed using scikit-build-core settings:

python -m pip install \
    -Cbuild-dir=build \
    -e \
    . \
    --config-settings=cmake.build-type="Debug" \
    -vv \
    --no-deps \
    --no-build-isolation

GitHub

The repository for PDAL's Python extension is available at https://github.com/PDAL/python

Python support released independently from PDAL itself as of PDAL 1.7.

Usage

Simple

Given the following pipeline, which simply reads an ASPRS LAS file and sorts it by the X dimension:

json = """
{
  "pipeline": [
    "1.2-with-color.las",
    {
        "type": "filters.sort",
        "dimension": "X"
    }
  ]
}"""

import pdal
pipeline = pdal.Pipeline(json)
count = pipeline.execute()
arrays = pipeline.arrays
metadata = pipeline.metadata
log = pipeline.log

Programmatic Pipeline Construction

The previous example specified the pipeline as a JSON string. Alternatively, a pipeline can be constructed by creating Stage instances and piping them together. For example, the previous pipeline can be specified as:

pipeline = pdal.Reader("1.2-with-color.las") | pdal.Filter.sort(dimension="X")

Stage Objects

• A stage is an instance of pdal.Reader, pdal.Filter or pdal.Writer.
• A stage can be instantiated by passing as keyword arguments the options applicable to the respective PDAL stage. For more on PDAL stages and their options, check the PDAL documentation on Stage Objects.
  • The filename option of Readers and Writers as well as the type option of Filters can be passed positionally as the first argument.
  • The inputs option specifies a sequence of stages to be set as input to the current stage. Each input can be either the string tag of another stage, or the Stage instance itself.
• The Reader, Filter and Writer classes come with static methods for all the respective PDAL drivers. For example, pdal.Filter.head() is a shortcut for pdal.Filter(type="filters.head"). These methods are auto-generated by introspecting pdal and the available options are included in each method's docstring:

>>> help(pdal.Filter.head)
Help on function head in module pdal.pipeline:

head(**kwargs)
    Return N points from beginning of the point cloud.

    user_data: User JSON
    log: Debug output filename
    option_file: File from which to read additional options
    where: Expression describing points to be passed to this filter
    where_merge='auto': If 'where' option is set, describes how skipped points should be merged with kept points in standard mode.
    count='10': Number of points to return from beginning.  If 'invert' is true, number of points to drop from the beginning.
    invert='false': If true, 'count' specifies the number of points to skip from the beginning.

Pipeline Objects

A pdal.Pipeline instance can be created from:

• a JSON string: Pipeline(json_string)
• a sequence of Stage instances: Pipeline([stage1, stage2])
• a single Stage with the Stage.pipeline method: stage.pipeline()
• nothing: Pipeline() creates a pipeline with no stages.
• joining Stage and/or other Pipeline instances together with the pipe operator (|):
  • stage1 | stage2
  • stage1 | pipeline1
  • pipeline1 | stage1
  • pipeline1 | pipeline2

Every application of the pipe operator creates a new Pipeline instance. To update an existing Pipeline use the respective in-place pipe operator (|=):

# update pipeline in-place
pipeline = pdal.Pipeline()
pipeline |= stage
pipeline |= pipeline2

Reading using Numpy Arrays

The following more complex scenario demonstrates the full cycling between PDAL and Python:

• Read a small testfile from GitHub into a Numpy array
• Filters the array with Numpy for Intensity
• Pass the filtered array to PDAL to be filtered again
• Write the final filtered array to a LAS file and a TileDB array via the TileDB-PDAL integration using the TileDB writer plugin

import pdal

data = "https://github.com/PDAL/PDAL/blob/master/test/data/las/1.2-with-color.las?raw=true"

pipeline = pdal.Reader.las(filename=data).pipeline()
print(pipeline.execute())  # 1065 points

# Get the data from the first array
# [array([(637012.24, 849028.31, 431.66, 143, 1,
# 1, 1, 0, 1,  -9., 132, 7326, 245380.78254963,  68,  77,  88),
# dtype=[('X', '<f8'), ('Y', '<f8'), ('Z', '<f8'), ('Intensity', '<u2'),
# ('ReturnNumber', 'u1'), ('NumberOfReturns', 'u1'), ('ScanDirectionFlag', 'u1'),
# ('EdgeOfFlightLine', 'u1'), ('Classification', 'u1'), ('ScanAngleRank', '<f4'),
# ('UserData', 'u1'), ('PointSourceId', '<u2'),
# ('GpsTime', '<f8'), ('Red', '<u2'), ('Green', '<u2'), ('Blue', '<u2')])
arr = pipeline.arrays[0]

# Filter out entries that have intensity < 50
intensity = arr[arr["Intensity"] > 30]
print(len(intensity))  # 704 points

# Now use pdal to clamp points that have intensity 100 <= v < 300
pipeline = pdal.Filter.expression(expression="Intensity >= 100 && Intensity < 300").pipeline(intensity)
print(pipeline.execute())  # 387 points
clamped = pipeline.arrays[0]

# Write our intensity data to a LAS file and a TileDB array. For TileDB it is
# recommended to use Hilbert ordering by default with geospatial point cloud data,
# which requires specifying a domain extent. This can be determined automatically
# from a stats filter that computes statistics about each dimension (min, max, etc.).
pipeline = pdal.Writer.las(
    filename="clamped.las",
    offset_x="auto",
    offset_y="auto",
    offset_z="auto",
    scale_x=0.01,
    scale_y=0.01,
    scale_z=0.01,
).pipeline(clamped)
pipeline |= pdal.Filter.stats() | pdal.Writer.tiledb(array_name="clamped")
print(pipeline.execute())  # 387 points

# Dump the TileDB array schema
import tiledb
with tiledb.open("clamped") as a:
    print(a.schema)

Reading using Numpy Arrays as buffers (advanced)

It's also possible to treat the Numpy arrays passed to PDAL as buffers that are iteratively populated through custom python functions during the execution of the pipeline.

This may be useful in cases where you want the reading of the input data to be handled in a streamable fashion, like for example:

• When the total Numpy array data wouldn't fit into memory.
• To initiate execution of a streamable PDAL pipeline while the input data is still being read.

To enable this mode, you just need to include the python populate function along with each corresponding Numpy array.

# Numpy array to be used as buffer
in_buffer = np.zeros(max_chunk_size, dtype=[("X", float), ("Y", float), ("Z", float)])

# The function to populate the buffer iteratively
def load_next_chunk() -> int:
"""
Function called by PDAL before reading the data from the buffer.

IMPORTANT: must return the total number of items to be read from the buffer.
The Pipeline execution will keep calling this function in a loop until 0 is returned.
"""
    #
    # Replace here with your code that populates the buffer and returns the number of elements to read
    #
    chunk_size = next_chunk.size
    in_buffer[:chunk_size]["X"] = next_chunk[:]["X"]
    in_buffer[:chunk_size]["Y"] = next_chunk[:]["Y"]
    in_buffer[:chunk_size]["Z"] = next_chunk[:]["Z"]

    return chunk_size

# Configure input array and handler during Pipeline initialization...
p = pdal.Pipeline(pipeline_json, arrays=[in_buffer], stream_handlers=[load_next_chunk])

# ...alternatively you can use the setter on an existing Pipeline
# p.inputs = [(in_buffer, load_next_chunk)]

The following snippet provides a simple example of how to use a Numpy array as buffer to support writing through PDAL with total control over the maximum amount of memory to use.

Example: Streaming the read and write of a very large LAZ file with low memory footprint

import numpy as np
import pdal

in_chunk_size = 10_000_000
in_pipeline = pdal.Reader.las(**{
    "filename": "in_test.laz"
}).pipeline()

in_pipeline_it = in_pipeline.iterator(in_chunk_size).__iter__()

out_chunk_size = 50_000_000
out_file = "out_test.laz"
out_pipeline = pdal.Writer.las(
    filename=out_file
).pipeline()

out_buffer = np.zeros(in_chunk_size, dtype=[("X", float), ("Y", float), ("Z", float)])

def load_next_chunk():
    try:
        next_chunk = next(in_pipeline_it)
    except StopIteration:
        # Stops the streaming
        return 0

    chunk_size = next_chunk.size
    out_buffer[:chunk_size]["X"] = next_chunk[:]["X"]
    out_buffer[:chunk_size]["Y"] = next_chunk[:]["Y"]
    out_buffer[:chunk_size]["Z"] = next_chunk[:]["Z"]

    print(f"Loaded next chunk -> {chunk_size}")

    return chunk_size

out_pipeline.inputs = [(out_buffer, load_next_chunk)]

out_pipeline.loglevel = 20 # INFO
count = out_pipeline.execute_streaming(out_chunk_size)

print(f"\nWROTE - {count}")

Executing Streamable Pipelines

Streamable pipelines (pipelines that consist exclusively of streamable PDAL stages) can be executed in streaming mode via Pipeline.iterator(). This returns an iterator object that yields Numpy arrays of up to chunk_size size (default=10000) at a time.

import pdal
pipeline = pdal.Reader("test/data/autzen-utm.las") | pdal.Filter.expression(expression="Intensity > 80 && Intensity < 120)")
for array in pipeline.iterator(chunk_size=500):
    print(len(array))
# or to concatenate all arrays into one
# full_array = np.concatenate(list(pipeline))

Pipeline.iterator() also takes an optional prefetch parameter (default=0) to allow prefetching up to to this number of arrays in parallel and buffering them until they are yielded to the caller.

If you just want to execute a streamable pipeline in streaming mode and don't need to access the data points (typically when the pipeline has Writer stage(s)), you can use the Pipeline.execute_streaming(chunk_size) method instead. This is functionally equivalent to sum(map(len, pipeline.iterator(chunk_size))) but more efficient as it avoids allocating and filling any arrays in memory.

Accessing Mesh Data

Some PDAL stages (for instance filters.delaunay) create TIN type mesh data.

This data can be accessed in Python using the Pipeline.meshes property, which returns a numpy.ndarray of shape (1,n) where n is the number of Triangles in the mesh.

If the PointView contains no mesh data, then n = 0.

Each Triangle is a tuple (A,B,C) where A, B and C are indices into the PointView identifying the point that is the vertex for the Triangle.

Meshio Integration

The meshes property provides the face data but is not easy to use as a mesh. Therefore, we have provided optional Integration into the Meshio library.

The pdal.Pipeline class provides the get_meshio(idx: int) -> meshio.Mesh method. This method creates a Mesh object from the PointView array and mesh properties.

Note

The meshio integration requires that meshio is installed (e.g. pip install meshio). If it is not, then the method fails with an informative RuntimeError.

Simple use of the functionality could be as follows:

import pdal

...
pl = pdal.Pipeline(pipeline)
pl.execute()

mesh = pl.get_meshio(0)
mesh.write('test.obj')

Advanced Mesh Use Case

USE-CASE : Take a LiDAR map, create a mesh from the ground points, split into tiles and store the tiles in PostGIS.

Note

Like Pipeline.arrays, Pipeline.meshes returns a list of numpy.ndarray to provide for the case where the output from a Pipeline is multiple PointViews

(example using 1.2-with-color.las and not doing the ground classification for clarity)

import pdal
import psycopg2
import io

pl = (
    pdal.Reader(".../python/test/data/1.2-with-color.las")
    | pdal.Filter.splitter(length=1000)
    | pdal.Filter.delaunay()
)
pl.execute()

conn = psycopg(%CONNNECTION_STRING%)
buffer = io.StringIO

for idx in range(len(pl.meshes)):
    m =  pl.get_meshio(idx)
    if m:
        m.write(buffer,  file_format = "wkt")
        with conn.cursor() as curr:
          curr.execute(
              "INSERT INTO %table-name% (mesh) VALUES (ST_GeomFromEWKT(%(ewkt)s)",
              { "ewkt": buffer.getvalue()}
          )

conn.commit()
conn.close()
buffer.close()

Requirements

• PDAL 2.6+
• Python >=3.9
• Pybind11 (eg pip install pybind11[global])
• Numpy >= 1.22 (eg pip install numpy)
• scikit-build-core (eg pip install scikit-build-core)