About

What is PDAL?

PDAL is Point Data Abstraction Library. It is a C/C++ open source library and applications for translating and processing point cloud data. It is not limited to LiDAR data, although the focus and impetus for many of the tools in the library have their origins in LiDAR.

What is its big idea?

PDAL allows you to compose operations on point clouds into pipelines of stages. These pipelines can be written in a declarative JSON syntax or constructed using the available API.

Why would you want to do that?

A task might be to load some ASPRS LAS (the most common LiDAR binary format) data into a database, but you wanted to transform it into a common coordinate system along the way.

One option would be to write a specialized monolithic program that reads LAS data, reprojects it as necessary, and then handles the necessary operations to insert the data in the appropriate format in the database. This approach has a distinct disadvantage in that without careful planning it could quickly spiral out of control as you add new little tweaks and features to the operation. It ends up being very specific, and it does not allow you to easily reuse the component that reads the LAS data separately from the component that transforms the data.

The PDAL approach is to chain together a set of components, each of which encapsulates specific functionality. The components allow for reuse, composition, and separation of concerns. PDAL views point cloud processing operations as a pipeline composed as a series of stages. You might have a simple pipeline composed of a LAS Reader stage, a Reprojection stage, and a PostgreSQL Writer, for example. Rather than writing a single, monolithic specialized program to perform this operation, you can dynamically compose it as a sequence of steps or operations.

_images/las-reproject-pgpointcloud.png — A simple PDAL pipeline composed of a reader, filter, and writer stages.

A PDAL JSON Pipeline that composes this operation to reproject and load the data into PostgreSQL might look something like the following:

{
  "pipeline":[
    {
      "type":"readers.las",
      "filename":"input.las"
    },
    {
      "type":"filters.reprojection",
      "out_srs":"EPSG:3857"
    },
    {
      "type":"writers.pgpointcloud",
      "connection":"host='localhost' dbname='lidar' user='hobu'",
      "table":"output",
      "srid":"3857"
    }
  ]
}

PDAL can compose intermediate stages for operations such as filtering, clipping, tiling, transforming into a processing pipeline and reuse as necessary. It allows you to define these pipelines as JSON, and it provides a command, pipeline, to allow you to execute them.

Note

Raster processing tools often compose operations with this approach. PDAL conceptually steals its pipeline modeling from GDAL’s Virtual Raster Format.

How is it different than other tools?

LAStools

One of the most common open source processing tool suites available for LiDAR processing is LAStools from Martin Isenburg. PDAL is different in philosophy in a number of important ways:

All components of PDAL are released as open source software under an OSI-approved license.
PDAL allows application developers to provide proprietary extensions that act as stages in processing pipelines. These might be things like custom format readers, specialized exploitation algorithms, or entire processing pipelines.
PDAL can operate on point cloud data of any format – not just ASPRS LAS. LAStools can read and write formats other than LAS, but relates all data to its internal handling of LAS data, limiting it to dimension types provided by the LAS format.
PDAL is coordinated by users with its declarative JSON syntax. LAStools is coordinated by linking lots of small, specialized command line utilities together with intricate arguments.
PDAL is an open source project, with all of its development activities available online at https://github.com/PDAL/PDAL

PCL

PCL is a complementary, rather than substitute, open source software processing suite for point cloud data. The developer community of the PCL library is focused on algorithm development, robotic and computer vision, and real-time laser scanner processing. PDAL can read and write PCL’s PCD format.

Entwine

Potree

Potree is a WebGL HTML5 point cloud renderer that speaks ASPRS LAS and LASzip compressed LAS. You can find the software at https://github.com/potree/potree/

Note

See Potree in action using the USGS 3DEP AWS Public Dataset at https://usgs.entwine.io

Others

Other open source point cloud softwares tend to be Desktop GUI, rather than library, focused. They include some processing operations, and sometimes they even embed tools such as PDAL. We’re obviously biased toward PDAL, but you might find useful bits of functionality in them. These other tools include:

Note

The libLAS project is an open source project that predates PDAL, and provides some of the processing capabilities provided by PDAL. It is currently in maintenance mode due to its dependence on LAS, the release of relevant LAStools capabilities as open source, and the completion of Python LAS software.

Where did PDAL come from?

PDAL takes its cue from another very popular open source project – GDAL. GDAL is Geospatial Data Abstraction Library, and it is used throughout the geospatial software industry to provide translation and processing support for a variety of raster and vector formats. PDAL provides the same capability for point cloud data types.

PDAL evolved out of the development of database storage and access capabilities for the U.S. Army Corps of Engineers CRREL GRiD project. Functionality that was creeping into libLAS was pulled into a new library, and it was designed from the ground up to mimic successful extract, transform, and load libraries in the geospatial software domain. PDAL has steadily attracted more contributors as other software developers use it to provide point cloud data translation and processing capability to their software.

How is point cloud data different than raster or vector geo data?

Point cloud data are indeed very much like the typical vector point data type of which many geospatial practitioners are familiar, but their volume causes some significant challenges. Besides their X, Y, and Z locations, each point often has full attribute information of other things like Intensity, Time, Red, Green, and Blue.

Typical vector coverages of point data might max out at a million or so features. Point clouds quickly get into the billions and even trillions, and because of this specialized processing and management techniques must be used to handle so much data efficiently.

The algorithms used to extract and exploit point cloud data are also significantly different than typical vector GIS work flows, and data organization is extremely important to be able to efficiently leverage the available computing. These characteristics demand a library oriented toward these approaches and PDAL achieves it.

Note

Possible point cloud dimension types provided and supported by PDAL can be found at Dimensions.

What tasks are PDAL good at?

PDAL is great at point cloud data translation work flows. It allows users to apply algorithms to data by providing an abstract API to the content – freeing users from worrying about many data format issues. PDAL’s format-free worry does come with a bit of overhead cost. In most cases this is not significant, but for specific processing work flows with specific data, specialized tools will certainly outperform it.

In exchange for possible performance penalty or data model impedance, developers get the freedom to access data over an abstract API, a multitude of algorithms to apply to data within easy reach, and the most complete set of point cloud format drivers in the industry. PDAL also provides a straightforward command line, and it extends simple generic Python processing through Numpy. These features make it attractive to software developers, data managers, and scientists.

What are PDAL’s weak points?

PDAL doesn’t provide a friendly GUI interface, it expects that you have the confidence to dig into the options of Filters, Readers, and Writers. We sometimes forget that you don’t always want to read source code to figure out how things work. PDAL is an open source project in active development, and because of that, we’re always working to improve it. Please visit Community to find out how you can participate if you are interested. The project is always looking for contribution, and the mailing list is the place to ask for help if you are stuck.

High Level Overview

PDAL is first and foremost a software library. A successful software library must meet the needs of software developers who use it to provide its software capabilities to their own software. In addition to its use as a software library, PDAL provides some command line applications users can leverage to conveniently translate, filter, and process data with PDAL. Finally, PDAL provides Python support in the form of embedded operations and Python extensions.

Core C++ Software Library

PDAL provides a C++ API software developers can use to provide point cloud processing capabilities in their own software. PDAL is cross-platform C++, and it can compile and run on Linux, OS X, and Windows. The best place to learn how to use PDAL’s C++ API is the test suite and its source code.

Command Line Utilities

PDAL provides a number of applications that allow users to coordinate and construct point cloud processing work flows. Some key tasks users can achieve with these applications include:

Print info about a data set
Data translation from one point cloud format to another
Application of exploitation algorithms
- Generate a DTM
- Remove noise
- Reproject from one coordinate system to another
- Classify points as ground/not ground
Merge or split data
Catalog collections of data

Note

The command line utilities are often simply pipeline and Pipeline collected into a convenient application. In many cases you can replicate the functionality of an application entirely within a single pipeline.

Python API

PDAL supports both embedding Python and extending with Python. These allow you to dynamically interact with point cloud data in a more comfortable and familiar language environment for geospatial practitioners.

Julia Plugin

PDAL supports embedding |Julia| filters. These allow you to dynamically interact with point cloud data in a more comfortable and familiar language environment for geospatial practitioners, while still maintaining high performance.

Additionally the TypedTables.jl, RoamesGeometry.jl and AcceleratedArrays.jl libraries provide some very high-level interfaces for writing efficient filters.

Conclusion

PDAL is an open source project for translating, filtering, and processing point cloud data. It provides a C++ API, command line utilities, and Python extensions. There are many open source software projects for interacting with point cloud data, and PDAL’s niche is in processing, translation, and automation.