Embodied AI Glossary中文

Stereo Baseline

基线Advanced

The distance between a stereo camera’s two lens centers, which sets how far and how accurately it can measure depth.

The baseline is the distance between the optical centers of a stereo camera’s left and right lenses. The stereo depth formula is depth = focal length × baseline / disparity, where disparity is the horizontal pixel difference between the same point as seen in the left and right images. A longer baseline gives more disparity at a given depth, so far-away measurements are more accurate, but the close-range blind zone grows and the overlap between the left and right images shrinks; a short baseline suits close range instead. So a camera should be chosen by baseline to match its working distance: wrist cameras usually use a short baseline, while mobile robots looking far ahead use a longer one. Depth error grows roughly with the square of distance and is inversely proportional to the baseline. In structured-light and active-stereo cameras, the distance between the projector and the camera plays a similar role.

ExampleThe ZED 2i has a baseline of about 12 cm, suited to scenes a few meters to over ten meters away; the wrist-mounted RealSense D405 has a very short baseline, suited to close range within tens of centimeters.

Also called
Baseline
Related
Stereo Camera · Disparity · Stereo Matching · Triangulation · Camera Extrinsics · Depth Camera
Sources
Depth Map from Stereo Images - OpenCV

See it in the full glossary →