Why machines use stereo vision at all
A single camera can't measure distance directly — it only sees a flat projection. Two cameras spaced apart, just like human eyes, let a machine calculate real-world distance to every point in a scene through triangulation, the same math behind human depth perception.
How it compares to LiDAR
Stereo camera rigs are far cheaper than LiDAR and work well in good lighting, but degrade in low light, fog or on featureless surfaces where the two images have nothing distinct to match. LiDAR is more consistent but far more expensive — many systems combine both.
Where it's used today
- Drone obstacle avoidance during autonomous flight
- Warehouse and delivery robot navigation
- Autonomous vehicle depth perception, often alongside radar and LiDAR
- Industrial robotic arms performing precision pick-and-place tasks
The connection back to photography
The underlying math — matching corresponding points between two offset views to calculate depth — is identical to what happens when a human brain fuses a stereoscopic photo pair. Robotics just automated the matching step.
See it in action at 3dstreaming.org