TY - GEN
T1 - An Efficient and Consistent Solution to the PnP Problem
AU - Zhou, Xiaoyan
AU - Xie, Zhengfeng
AU - Yu, Qida
AU - Zong, Yuan
AU - Wang, Yiru
N1 - Publisher Copyright:
© 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2024
Y1 - 2024
N2 - In this paper, we present a novel non-iterative algorithm for solving the pose estimation problem from a set of 3D-to-2D point correspondences, known as the Perspective-n-Point (PnP) problem. The presented algorithm is capable of achieving both geometrical and statistical optimality by exploring the geometrical constraints of the PnP problem through a nonlinear least-squares fashion, as well as accounting for observation uncertainty in the solution process. In addition, to further improve the accuracy of the presented algorithm, we introduce a method that is able to eliminate the bias of solution caused by the propagation of uncertainty, resulting in a consistent estimate. Experimental tests on synthetic data and real images (i.e., TempleRing dataset) show that the presented algorithm can well adapt to different levels of noise, and out-perform state-of-the-art (SOTA) PnP algorithms in terms of accuracy and computational cost. This makes the presented algorithm eminently suitable for a wide range of application scenarios.
AB - In this paper, we present a novel non-iterative algorithm for solving the pose estimation problem from a set of 3D-to-2D point correspondences, known as the Perspective-n-Point (PnP) problem. The presented algorithm is capable of achieving both geometrical and statistical optimality by exploring the geometrical constraints of the PnP problem through a nonlinear least-squares fashion, as well as accounting for observation uncertainty in the solution process. In addition, to further improve the accuracy of the presented algorithm, we introduce a method that is able to eliminate the bias of solution caused by the propagation of uncertainty, resulting in a consistent estimate. Experimental tests on synthetic data and real images (i.e., TempleRing dataset) show that the presented algorithm can well adapt to different levels of noise, and out-perform state-of-the-art (SOTA) PnP algorithms in terms of accuracy and computational cost. This makes the presented algorithm eminently suitable for a wide range of application scenarios.
KW - PnP Problem
KW - Pose Estimation
KW - Statistically Optimal
UR - http://www.scopus.com/inward/record.url?scp=85180801424&partnerID=8YFLogxK
U2 - 10.1007/978-981-99-8432-9_17
DO - 10.1007/978-981-99-8432-9_17
M3 - Conference contribution
AN - SCOPUS:85180801424
SN - 9789819984312
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 207
EP - 220
BT - Pattern Recognition and Computer Vision - 6th Chinese Conference, PRCV 2023, Proceedings
A2 - Liu, Qingshan
A2 - Wang, Hanzi
A2 - Ji, Rongrong
A2 - Ma, Zhanyu
A2 - Zheng, Weishi
A2 - Zha, Hongbin
A2 - Chen, Xilin
A2 - Wang, Liang
PB - Springer Science and Business Media Deutschland GmbH
T2 - 6th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2023
Y2 - 13 October 2023 through 15 October 2023
ER -