AprilRobotics/apriltag

What it solves

AprilTag provides a robust visual fiducial system that allows robots and computer vision systems to detect specific markers (tags) in an image. It solves the problem of identifying unique landmarks and estimating their 3D pose (position and orientation) relative to a camera, which is essential for navigation and localization in robotics.

How it works

The system uses a C library to detect specific patterns of black and white squares (tags) within an image. It identifies the quadrilateral border of the tag and then decodes the internal data bits to identify which specific tag from a family is present. The library supports various tag layouts, including standard, circular, and custom designs, and can estimate the tag's pose by using the camera's focal length and center coordinates combined with the known physical size of the tag.

Who it’s for

This tool is primarily for robotics researchers and developers building systems that require precise visual markers for localization, docking, or object tracking.

Highlights

• High Performance: AprilTag 3 features a detector that is over twice as fast as previous versions and has improved detection rates for small tags.
• Flexible Layouts: Supports diverse tag designs, including circular tags and layouts with holes for multi-distance detection.
• Pose Estimation: Includes built-in methods to calculate the 3D position and orientation of a tag relative to the camera.
• Minimal Dependencies: Written as a small C library with minimal external requirements, making it easy to integrate into various environments.