PuzzlePoles: Cylindrical Fiducial Markers Based on the PuzzleBoard Pattern
Juri Zach, Peer Stelldinger · 2025
Read PuzzlePoles: Cylindrical Fiducial Markers Based on the PuzzleBoard Pattern by Juri Zach, Peer Stelldinger (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-19448.txt.
Introduction
Juri Zach and Peer Stelldinger introduce PuzzlePoles, cylindrical visual markers that machines can read from any side. Autonomous systems rely on known marker patterns that cameras recognise for setup and location fixes, but flat markers fail when viewed edge-on or from behind. The PuzzlePole wraps the PuzzleBoard pattern, with its distinctive puzzle structure, around a pole, so recognition and pose estimation work across the full circle of viewing directions.
The pattern's uniqueness supports accurate location and orientation while tolerating partial blocking, since a visible slice can still identify the pole. The design adapts to many uses, from robot navigation and live mapping to tangible interfaces that people handle. The authors present the marker as a flexible addition to the sensing toolkit for autonomous systems.
Catalog record
| CREATOR | Juri Zach, Peer Stelldinger |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-19448
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
PuzzlePoles: Cylindrical Fiducial Markers Based on the PuzzleBoard Pattern Juri Zach, Peer Stelldinger Abstract Reliable perception of the environment is a key enabler for autonomous systems, where calibration and localization tasks often rely on robust visual markers. We introduce the PuzzlePole , a new type of fiducial markers derived from the recently proposed PuzzleBoard calibration pattern. The PuzzlePole is a cylindrical marker, enabling reliable recognition and pose estimation from 360 ∘ 360^{\circ} viewing direction. By leveraging the unique combinatorial structure of the PuzzleBoard pattern, PuzzlePoles provide a high accuracy in localization and orientation while being robust to occlusions. The design offers flexibility for deployment in diverse autonomous systems scenarios, ranging from robot navigation and SLAM to tangible interfaces. 1 Introduction and Previous Work Accurate and reliable perception of the environment is crucial for autonomous systems to perform tasks such as navigation, mapping, localization, and object interaction. Fiducial markers have been widely adopted as artificial landmarks that provide well-defined geometric features for camera-based pose estimation and system calibration. These markers facilitate robust and efficient extraction of 2D-3D correspondences needed for localization algorithms and have been instrumental in advancing visual perception capabilities. Among the most popular fiducial marker systems is ArUco [ 5 ] , which consists of square binary patterns that can be quickly detected and decoded using a dictionary-based approach. ArUco markers are favored for their simplicity, low computational requirements, and ability to work reliably under various illumination and partial occlusion conditions. However, ArUco markers are planar and typically inflexible in terms of large viewpoint variations, often suffering degradation in pose estimation accuracy when viewed at oblique angles or from multiple directions [ 11 ] . The same is true for similar planar fiducial markers like ARTag [ 3 ] or AprilTag [ 12 ] . For camera calibration purposes, Charuco patterns [ 9 ] extend the ArUco concept by combining chessboard and ArUco markers, providing improved corner detection and enhanced geometric constraints leading to higher calibration accuracy. ArUco grids and Charuco markers are well-suited for camera calibration but maintain the limitation of planarity and directionality inherent to their flat design. Beyond planar markers, 3D fiducial objects can potentially be detected from arbitrary directions. Beyond planar designs, 3D fiducial objects such as not only regular polyhedra [ 4 ] but also cylindrical markers have been investigated: While some cylindrical markers are invariant regarding the rotation around the cylinder axis [ 7 ] , others are restricted to a certain rotational sector [ 10 , 20 ] or allow to detect the rotation for the whole 360° range [ 18 ] . Wang et al. [ 18 ] place 12+ stripe-like ”feature subregions” around the circumference of a cylinder in a de Bruijn-sequence-like encoding [ 2 , 6 ] to define their CylinderTag. As at least two feature subregions need to be completely visible, their approach is not robust to partial occlusions. In contrast, Jayarathne et al. [ 10 ] and Zhang et al. [ 20 ] distribute local features on the cylinder surface to gain some robustness to partial occlusions. Both use chessboard corners as local features which Zhang et al. combine with circular dots. However, none of these approaches gives a 360° cylindrical encoding with occlusion robustness. A notable exception is an approach being presented just for comparison in the CylinderTag paper [ 18 ] : Wrapping a grid of standard 2D AprilTag fiducial markers around a cylinder obviously solves this task, but at the cost of the need of a high image resolution as a large number of individual tags needs to be decodable. Similarly, Xiao et al. present a 2d pattern which can be wrapped around a cylinder for optical tracking for tangible interaction and compare this to 2D ArUco markers on a cylinder surface [ 19 ] . Again their solution, as well as the ArUco alternative, is robust to occlusions but requires some quite high resolution. In order to be more robust regarding both occlusions and low resolution, one can distribute the local position encoding in overlapping areas by using so-called marker fields [ 17 ] or de Bruijn tori [ 8 , 13 , 15 ] . Such an approach has been proposed by Zhu et al. [ 21 ] , which has also been applied to cylindrical surfaces already [ 1 ] . A similar and even more robust pattern is the recently proposed PuzzleBoard calibration pattern [ 14 ] . The PuzzleBoard pattern presents a novel combinatorial design which integrates de Bruijn rings [ 15 ] into a chessboard pattern. It permits high spatial encoding capacity and robust corner extraction. The PuzzleBoard pattern has been shown to outperform classical calibration patterns in terms of detection robustness and calibration accuracy both in case of severe occlusions and low image resolutions due to its ability to resolve ambiguities and provide rich spatial information. 2 PuzzlePole: A Cylindrical Marker We leverage the PuzzleBoard pattern to introduce a new fiducial marker type: PuzzlePole , a cylindrical marker utilizing cyclic PuzzleBoard subpatterns to enable 360^{\circ} recognition and pose estimation. By using the combinatorial strength and spatial richness of the PuzzleBoard pattern we aim to address the shortcomings of existing cylindrical fiducial marker approaches. PuzzlePole markers are cylindrical poles wrapped with cyclic PuzzleBoard subpatterns. This configuration enables reliable detection and pose estimation from multiple viewing angles and directions, overcoming occlusions and viewpoint challenges typical for planar markers. The main advantage of the PuzzleBoard pattern compared to a grid of ArUco-like fiducial markers is that the position encoding of local patches is distributed on neighboring base cells, which allows decodability at much lower resolutions [ 14 ] : Each patch of 3x3 puzzle pieces decodes a unique position by the configuration of its tabs and knobs. To identify tabs and knobs at low resolutions it is sufficient to check if the center between two corner points is black or white. However, wrapping such a pattern around a cylinder in a seamless way requires periodicity of the code. Otherwise the code would be destroyed near the seam. Although the original PuzzleBoard pattern is periodic, it repeats only every 501 grid cells, which is nice for calibration purposes but disadvantageous when being used on a cylinder. Figure 1: A sub-pattern of the PuzzleBoard pattern, which starts at y=73 (and x=0 ). As can be seen, the last row of puzzle pieces is identical to the first one. By wrapping the pattern onto a vertical cylinder and placing the first and last row exactly onto each other, one gets a seamless cyclic pattern. As a solution, one could generate new de Bruijn-like patterns with the appropriate period length, by using random uniform marker fields as in [ 1 , 6 , 17 , 21 ] . We choose another approach: We use quasiperiodic sub-patterns of the PuzzleBoard pattern, as this gives us a simple decoding algorithm with efficient error correction. Quasiperiodicity here means that the sub-pattern can be repeated twice within the whole pattern without generating new local 3-by-3 codes at the seams. As the PuzzleBoard pattern, as defined by Stelldinger et al. [ 14 ] , is based on a periodic overlay of a horizontal code stripe of width 3 and