Solving the PnL problem using the hidden variable method: an accurate and efficient solution

• Published in: The Visual computer Vol. 38; no. 1; pp. 95 - 106
• Main Authors: Wang, Ping, Chou, Yongxin, An, Aimin, Xu, Guili
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 2022; Springer Nature B.V
• Subjects: Accuracy; Artificial Intelligence; Cameras; Computer Graphics; Computer Science; Computing costs; Coordinate transformations; Efficiency; Image Processing and Computer Vision; Iterative methods; Maximum likelihood method; Newton methods; Original Article; Parameterization; Polynomials; Pose estimation; Computer vision; Camera pose estimation; Perspective-n-line problem (PnL); Absolute position and orientation
• ISSN: 0178-2789; 1432-2315
• DOI: 10.1007/s00371-020-02004-2

This paper addresses the camera pose estimation problem from 3D lines and their 2D projections, known as the perspective-n-line (PnL) problem. Although many successful solutions have been presented, it is still a challenging to optimize both computational complexity and accuracy at the same time. In our work, we parameterize the rotation by using the Cayley–Gibbs–Rodriguez (CGR) parameterization and formulate the PnL problem into a polynomial system solving problem. Instead of the Gröbner basis method, which may encounter numeric problems, we seek for an efficient and stability technique—the hidden variable method—to solve the polynomial system and polish the solution via the Gauss–Newton method. The performance of our method is evaluated by using simulations and real images, and results demonstrate that our method offers accuracy and precision comparable or better than existing state-of-the-art methods, but with significantly lower computational cost.