EUVIMEDEUVIMED entdecken
Crossref

EA-GPnP: Efficient and Accurate Generalized-Perspective-n-Point Solution via Optimized Null Space Analysis

Yi Zhang, Baoqiong Wang, Kunhong Li, Wenjun Chen, Xiuqi Wang, Ye Zhang, Yueqiang Zhang, Yulan Guo, Qifeng Yu

IEEE Transactions on Robotics · 2026 · 42 · 2541-2554

DOI: 10.1109/tro.2026.3701596 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Robotics, 2026. Verfasst von Yi Zhang und 8 weiteren Autor:innen. Die Arbeit nennt 67 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Basic and Applied Basic Research Foundation of Guangdong Province, Science and Technology Planning Project of Key Laboratory of Advanced IntelliSense Technology.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In recent years, multicamera systems have been increasingly popular in various robotic tasks due to the growing demand for wider fields of view. A key challenge is to locate the system accurately and efficiently. The common approach involves solving the generalized perspective-n-points (gPnP) problem, which estimates the pose of a calibrated multicamera system using 3D–2D point correspondences. Unfortunately, existing methods still face a tradeoff between accuracy and efficiency. Specifically, the efficient null-space-analysis-based (NSA-based) method compromises accuracy due to its suboptimal NSA framework, while the accurate optimality-based methods incur heavy computational loads from handling sophisticated polynomial systems. To bridge this gap, we propose EA-GPnP, a novel NSA-based gPnP method that achieves superior accuracy and further enhances efficiency. The core of our approach consists of three targeted optimizations to the NSA framework, concerning its formulation, overall structure, and polynomial solving. In addition, we develop two specialized solvers tailored for polynomials with 6 quadratic equations in 3 or 4 unknowns (6Q3 or 6Q4), yielding up to four or nine solutions, respectively. Extensive real and synthetic experiments demonstrate that our method not only outperforms existing methods in numerical performance and noise resilience but also achieves the fastest speed. It achieves up to a 3.5× speedup over the most accurate state-of-the-art method, while concurrently delivering better performance.

Förderung

  • National Natural Science Foundation of China
  • Basic and Applied Basic Research Foundation of Guangdong Province
  • Science and Technology Planning Project of Key Laboratory of Advanced IntelliSense Technology
  • Guangdong Science and Technology Department
  • Aeronautical Science Fund of China

Zitieren

APA

Zhang, Y., Wang, B., Li, K., Chen, W., Wang, X., Zhang, Y., Zhang, Y., Guo, Y., & Yu, Q. (2026). EA-GPnP: Efficient and Accurate Generalized-Perspective-n-Point Solution via Optimized Null Space Analysis. IEEE Transactions on Robotics, 42, 2541-2554. https://doi.org/10.1109/tro.2026.3701596

Vancouver

Zhang Y, Wang B, Li K, Chen W, Wang X, Zhang Y, et al. EA-GPnP: Efficient and Accurate Generalized-Perspective-n-Point Solution via Optimized Null Space Analysis. IEEE Transactions on Robotics. 2026;42:2541-2554. doi:10.1109/tro.2026.3701596

BibTeX
@article{Zhang2026,
  author = {Zhang, Yi and Wang, Baoqiong and Li, Kunhong and Chen, Wenjun and Wang, Xiuqi and Zhang, Ye and Zhang, Yueqiang and Guo, Yulan and Yu, Qifeng},
  title = {EA-GPnP: Efficient and Accurate Generalized-Perspective-n-Point Solution via Optimized Null Space Analysis},
  journal = {IEEE Transactions on Robotics},
  year = {2026},
  volume = {42},
  pages = {2541--2554},
  doi = {10.1109/tro.2026.3701596}
}
Kostenlos starten →