Abdallah Bengueddoudj¹ Foudil Belhadj¹ Yongtao Hu² Brahim Zitouni³ Yacine Idir¹ Ibtissem Adoui¹ Messaoud Mostefai¹

¹ Mohamed El-Bachir El-Ibrahimi University of Bordj Bou Arreridj, Algeria      ² Guangdong Virtual Reality Technology Co., Ltd., Guangzhou, China      ³ Private Optical Institute of Bordj Bou Arreridj, Algeria

Informatica (Informatica 2025)

---

Abstract

Today, the most widely used visual markers, such as ArUco and AprilTag, rely on square pixel arrays. While these markers can deliver satisfactory detection and identification outcomes, they remain vulnerable to corner occlusion despite incorporating corrective codes. Conversely, line-based markers offer increased resilience against occlusions but are typically constrained in terms of codification capacities. The markers developed in this research leverage linear information to propose a pyramidal line-based structure that exhibits robustness to corner occlusion while providing enhanced coding capacities. Moreover, the projective invariance of the constituent lines enables the validation of a homography-less identification method that considerably reduces computation resources and processing time. We assembled an extensive test dataset of 169,713 images for evaluation, including rotation, distances, and different levels of occlusion. Experiments on this dataset show that the proposed marker significantly outperforms previous fiducial marker systems across multiple metrics, including execution time and detection performance under occlusion. It effectively identifies markers with up to 50% occlusion and achieves identification at a resolution of 1920x1080 in 17.20 ms. The developed marker generation and identification, as well as an extensive marker Database, are publicly available for tests at https://github.com/OILUproject/OILUtag

Downloads

• Paper (PDF, 1.2 MB)

Bibtex

@article{Bengueddoudj2025oilutag,
  title={{Efficient Line-Based Visual Marker System Design with Occlusion Resilience}},
  author={Bengueddoudj, Abdallah and Belhadj, Foudil and Hu, Yongtao and Zitouni, Brahim and Idir, Yacine and Adoui, Ibtissem and Mostefai, Messaoud},
  journal={Informatica},
  volume={49},
  number={1},
  pages={61-80},
  year={2025},
  publisher={}
}