Breaking the Tradeoff Between Accuracy and Distance: A Long-Focal Camera Pose Measurement Method and Verification System

Long-distance, high-precision attitude measurement for large spatial spacecraft has become a cornerstone of successful space missions. While long-focal-length cameras are essential for addressing long-distance measurement challenges, conventional perspective-$n$-point (PnP) methods suffer from rank deficiency in the coefficient matrix during linear solving and severe singularity of the Hessian matrix in nonlinear optimization, rendering ultraprecise measurement extremely difficult. To address these limitations, we propose LFPnP, a novel PnP framework specifically tailored for long-distance, long-focal-length cameras. In the linear phase, we introduce a vectorized joint-solving strategy based on a weak perspective model, achieving global synchronization that effectively eliminates the linear dependence in the coefficient matrix. In the nonlinear phase, we construct a constant-damping regularized optimization problem in the Lie algebra space [$\mathfrak {se}(3)$], leveraging anchored damping factors to fix the singularity of the Hessian matrix, thereby significantly reducing geometric approximation errors and suppressing divergence. Furthermore, to overcome the challenge of acquiring absolute ground truth at long distances, we establish a metrology-grade precision verification system based on a “camera–target–laser tracker” setup. Experimental results demonstrate that LFPnP consistently outperforms state-of-the-art methods across all dimensions. Specifically, at an ultralong distance of 60 m, LFPnP constrains the relative attitude errors along the$x$-,$y$-, and$z$-axes to merely$2.40"$,$3.15"$, and$2.36"$, respectively, while keeping the relative translation errors within 22.64, 35.65, and$665.68~\mu \text {m}$, demonstrating significant superiority over existing SOTA methods. The code is publicly available athttps://anonymous.4open.science/r/LFPnP_MATLAB/LFPnP_MATLAB.zip