Calibrating a Multi-Camera Robot Head

Multiple cameras only work together if the system knows exactly how they're positioned relative to each other.

Combining data from several cameras (or a camera with a depth sensor) requires knowing the precise geometric relationship between them — their relative position and orientation — a process called extrinsic calibration, usually done alongside intrinsic calibration (correcting each individual camera's own lens distortion and focal properties).

How it's typically done

A checkerboard or similar known-pattern target is shown to all the cameras simultaneously from various angles, and calibration software solves for the camera parameters that best explain how that same pattern appears differently in each view.

Why it needs to be redone

Physical shocks, temperature changes, and general wear can shift a camera's mounting by fractions of a millimeter — enough to meaningfully degrade depth accuracy over time. Production robots often re-verify calibration periodically or after any significant impact, rather than assuming a one-time factory calibration holds forever.