ASCII
Human readable, easy to diff and inspect, and considerably larger and slower to load. Fine for a few thousand points.
PCD stands for Point Cloud Data. It is the native file format of the Point Cloud Library (PCL), and it is what most LiDAR and SLAM tooling in robotics writes out by default.
A PCD file is a plain-text header followed by point data. The header declares the field layout, so a reader knows what each number means before parsing anything. A minimal LiDAR cloud carries x, y and z; recordings often add intensity, and some drivers add ring, timestamp or RGB fields.
# .PCD v0.7 - Point Cloud Data file format
VERSION 0.7
FIELDS x y z intensity
SIZE 4 4 4 4
TYPE F F F F
COUNT 1 1 1 1
WIDTH 1250000
HEIGHT 1
VIEWPOINT 0 0 0 1 0 0 0
POINTS 1250000
DATA binaryHuman readable, easy to diff and inspect, and considerably larger and slower to load. Fine for a few thousand points.
Raw point records after the header. The usual choice for mapped clouds, which routinely run into tens of millions of points.
Same layout with LZF compression applied. Smaller on disk, and not every third-party reader supports it.
Three common routes: a LiDAR vendor SDK exporting a single scan, a PCL command line tool converting some other capture, or a SLAM pipeline writing out the finished map. The last one is the interesting case, because a mapped PCD is not just a dump of scans stacked on top of each other. It is the result of registering every scan into one consistent frame, closing loops, and correcting drift.
That is the path LidarFlow automates: upload a .bag or .mcap recording, and the run returns output.pcd alongside a PLY export and the sensor trajectory. No PCL build, no SLAM parameter tuning.
PLY is the general-purpose 3D format that also carries meshes. LAS and its compressed sibling LAZ dominate survey and GIS work. E57 is the interoperability format for terrestrial scanning. Rather than half-explaining each here, see the full comparison on point cloud file formats.
PCL-based tools and CloudCompare read PCD directly, and ROS/RViz can display one. In the browser you can open it with the free LidarFlow PCD viewer, which parses the file locally in the tab.
PCD comes from the Point Cloud Library and stores a named field layout per point, which makes it natural for LiDAR data with intensity or ring fields. PLY is a general 3D format that also stores faces, so it is better understood by mesh and graphics software.
Yes. The free point cloud converter handles PCD, PLY, XYZ and CSV in the browser without uploading anything. For LAS or LAZ, CloudCompare and PDAL are the usual desktop route.
Running SLAM over the recording produces one. That can be a local PCL or GLIM setup, or you can upload a .bag or .mcap to LidarFlow and get output.pcd back without installing ROS.