Point clouds appeared in the robotics toolbox as a way to intuitively represent and manipulate the information provided by 3D sensors, such as time-of-flight cameras and laser scanners, in which space is sampled in a finite set of points in a 3D frame of reference. The Point Cloud Library (PCL) provides a number of data types and data structures to easily represent not only the points of our sampled space, but also the different properties of the sampled space, such as color and normal vectors. PCL also provides a number of state-of-the-art algorithms to perform data processing on our data samples, such as filtering, model estimation, and surface reconstruction.

ROS provides a message-based interface through which PCL point clouds can be efficiently communicated, and a set of conversion functions from native PCL types to ROS messages, in much the same way as it is done with OpenCV images. Aside from the standard capabilities of the ROS API, there are a number of standard packages that can be used to interact with common 3D sensors, such as the widely used Microsoft Kinect or the Hokuyo laser, and visualize the data in different reference frames with the RViz visualizer.

This chapter will provide a background on the PCL, relevant data types, and ROS interface messages that will be used throughout the rest of the sections. Later, a number of techniques will be presented on how to perform data processing using the PCL library and how to communicate the incoming and outgoing data through ROS.