Embodied AI Glossary中文

AprilTag

Common

A printable black-and-white square marker that a camera can detect to compute the marker's ID and 6D pose.

AprilTag is a visual fiducial marker system proposed by Edwin Olson's team (the APRIL lab) at the University of Michigan, presented at ICRA 2011, and it looks like a simplified QR code. It only encodes a single ID number rather than large amounts of data, but the detector reliably finds its four corners, and combining those with the camera's intrinsics and the tag's known physical size, PnP gives its 6D pose relative to the camera. In robotics it's commonly used for hand-eye calibration, providing ground-truth poses for objects or worktables, and aligning the extrinsics of multiple cameras. The current open-source version, AprilTag 3 (a C implementation under the BSD license), offers several tag families such as tag36h11 and tagStandard41h12, and can also detect ArUco markers.

ExamplePasting tag36h11 markers at the four corners of a worktable lets a fixed third-person camera detect them and solve for the table's pose relative to the camera, which is then used to bring several cameras into one shared coordinate frame.

Also called
Visual Fiducial Marker, Fiducial Marker, AprilTag 3
Related
ArUco Marker · Hand-Eye Calibration · Perspective-n-Point · Camera Extrinsics · 6D Object Pose Estimation · Calibration Board
Sources
AprilTag(University of Michigan APRIL Lab)
AprilRobotics/apriltag(AprilTag 3)

See it in the full glossary →