3D bounding box annotation is a method of marking targets with cuboid boxes in 3D data (e.g., point clouds, 3D models), including center coordinates, length, width, height and rotation angle, clarifying target spatial position and size, used for 3D object detection model training.