Reason for using pcd format

In most AV datasets (Nuscenses , Apollo and Ford) lidar dataset is in point cloud data (pcd) format. For example, Nuscenes data is in binary pcd format whereas others are in binary compressed format.

Is pcd standard for lidar data. I am curious as to what is the format in which lidar equipment dumps the data while test run. Why not using LAS format which was common for GIS systems ?

Pcd is a simple and open pointcloud format that is not application specific. You just specify the name and data types of your points. LAS (https://www.loc.gov/preservation/digital/formats/fdd/fdd000418.shtml) seems to be a much more application specific format as shown below. I suppose it wouldn’t be very difficult to convert the formats, but most of the fields would take dummy values.

  • : Core-0. 20 bytes. Includes x,y,z values, intensity of pulse return, return identification, classification (see below), scan angle, and identifier for the source for this point data, if derived from an earlier dataset.
  • 1: Core-0 + GPS time
  • 2: Core-0 + three 16-bit color channels
  • 3: Core-0 + GPS time and three 16-bit color channels
  • 4: Type 1 + wave packets. Wave packets may be stored in the file or in an auxiliary file with the same name and extension .wdp .
  • 5: Type 3 + wave packets
  • 6: Core-6 (allowing more bits for return data, and making GPS time mandatory)
  • 7: Core-6 + three 16-bit color channels
  • 8: Type 7 + NIR (near infra-red) channel
  • 9: Core-6 + wave packets
  • 10: Type 9 + three color channels and NIR channel.