Publications
WORKSHOP (INTERNATIONAL) GOOP-PnPL: Global optimization of orthogonal projection error based PnPL
Yasuyuki Sugaya (Toyohashi University of Technology), Fumiya Okamoto (Toyohashi University of Technology), Norio Kosaka, Sakiko Nishi, Kazuhiro Ninomiya
The IEEE/CVF Conference on Computer Vision and Pattern Recognition 2026 (CVPR 2026)
June 07, 2026
In this paper, we address the Perspective-n-Point-and-Line problem (PnPL), which estimates a camera pose from a mixture of point and line features with known 3-D and 2-D correspondences. Unlike existing methods, which use orthogonal projection errors, we first derive a unified formulation of these errors to deal with them together. Based on this formulation, we propose two complementary methods. We then refine the closed-form solution by applying our alternate method. In various experiments, we confirmed the efficiency of our proposed methods. We also compared our method with existing methods and confirmed that our method is superior to the EPnPL in terms of accuracy and to the OPnPL in terms of computation time.
Paper :
GOOP-PnPL: Global optimization of orthogonal projection error based PnPL
(external link)
PDF : GOOP-PnPL: Global optimization of orthogonal projection error based PnPL