ICOTAG3D: a 3D marker for enhanced six-degree-of-freedom pose estimation for robots

Applications such as robotic assembly, inspection, and localization in complex manufacturing environments within intelligent manufacturing rely heavily on pose estimation. The use of fiducial markers for highly accurate pose estimation is particularly crucial. However, the accuracy and detection capabilities of pose estimation are often constrained by the shape and scale of the fiducial markers. In this paper, we introduce IcoTag3D, a three-dimensional marker featuring triangles affixed to an icosahedron, designed specifically for pose estimation. This innovative design extends the angular observation range and increases the marker's scale, thereby enhancing estimation accuracy and recognition distance. Two-dimensional coordinates of the marker feature points are detected and extracted from the environment. The 3D coordinates of these feature points are then obtained using triangulation. This process creates 2D-3D point pairs. High-quality interior points are then filtered using the Random Sampling Consensus (RANSAC) method. The initial positions are determined by the Efficient PnP (EPnP) method and then Levenberg-Marquardt (LM) optimization is applied. We evaluate the performance of IcoTag3D through simulation and physical experiments. The results demonstrate the advantages and feasibility of IcoTag3D.