Point cloud
464 words Β· 3 min read Β· 2 sources
A point cloud is a collection of millions of individual 3D coordinates, each one marking a surface that a laser or depth camera measured, together forming a precise three-dimensional map of the environment.
The concept concept: A point cloud is a collection of millions
Difficulty 3/5 Β· ClassroomImagine pressing a piece of tracing paper against a brick wall and dotting every pore, every groove, every bump with a pencil. Lift the paper away and you have a flat scatter of dots that, together, describe the wall's surface with surprising accuracy. Now do the same thing in three dimensions β instead of a flat paper, a sphere of dots radiating outward fro
π‘ Think of it likeβ¦
Think of it like a household object that does the same job β the underlying idea is the same, just adapted for robots.
Why it matters
Without point cloud, many concept systems in robotics simply couldn't work.
Imagine pressing a piece of tracing paper against a brick wall and dotting every pore, every groove, every bump with a pencil. Lift the paper away and you have a flat scatter of dots that, together, describe the wall's surface with surprising accuracy. Now do the same thing in three dimensions β instead of a flat paper, a sphere of dots radiating outward from the sensor in every direction β and you have a rough idea of what a point cloud looks like.
A point cloud is a set of data points in three-dimensional space. Each point has an X, Y, and Z coordinate (and sometimes additional information like colour or intensity), and together the millions of points describe the surfaces of everything the sensor could see, like a 3D dot-to-dot picture of the real world.
Where point clouds come from
The most common source is lidar: a laser fires pulses in every direction, measures the distance to whatever each pulse hits, and records the 3D coordinates of each impact. A spinning automotive lidar like a Velodyne HDL-64 generates roughly 1.3 million points per second. Depth cameras β like the Intel RealSense or the structured-light sensor in the Microsoft Kinect β produce denser but shorter-range point clouds. Photogrammetry, the process of calculating depth from many overlapping photographs, is another source.
What robots do with a point cloud
Raw point clouds are too large and unstructured to act on directly. A standard pipeline processes them through several stages:
- Downsampling β reduce the number of points while preserving shape, to fit in memory and speed up computation.
- Segmentation β separate the cloud into clusters that correspond to individual objects.
- Registration β align two clouds taken from different positions to build a larger consistent map.
- Surface reconstruction β convert the cloud into a solid mesh that other algorithms can reason about.
A real example
When a construction company wants to document a partially-built structure, a surveyor walks the site with a handheld lidar scanner. The resulting point cloud captures every beam, pipe, and wall at millimetre accuracy. Engineers back at the office overlay this cloud on the original CAD model to see exactly where the construction deviates from the plan β work that used to take weeks of manual measurement now takes hours.
Why it matters for robots
A robot cannot grasp what it cannot locate in 3D space. Point clouds are often the richest 3D description a robot has of the objects around it, and algorithms that process point clouds are central to grasping, obstacle avoidance, and map building.
The same point-cloud techniques used by self-driving cars are being applied to ancient ruins β archaeologists now scan crumbling temples into digital 3D archives before they deteriorate further.
Ask R2 Co-pilot anything you didn't understand about Point cloud. It'll explain it plainly.
Keep going
A* (A-Star) Pathfinding in Robotics β Complete Guide
A* finds the shortest path between two points on a grid or graph. It is the most-used pathfinding algorithm inβ¦
ConceptAccelerometer in Robotics β Complete Guide
An accelerometer measures linear acceleration along an axis. In robotics, accelerometers detect motion, tilt, β¦
ConceptActuator
The muscles of a robot β devices that convert electrical or pneumatic energy into mechanical motion.
Last updated Β· 2026-05-19
Community discussion
0 questions & insightsLoading discussionβ¦
Spotted something off? Report an error β