Surface Pose Monocular Measurement Method Based on Center Bias Correction

To address pose estimation errors caused by the misalignment between the projected center of a curved-surface circular feature and its image centroid, this paper proposes a monocular pose correction method. A geometric model of the curved-surface feature is first established to describe the 3D coordinates of points on the surface, which are then projected onto the image plane using camera intrinsic and extrinsic parameters. Unlike contour-based approaches, the centroid of the projected feature is computed based on area integration, ensuring consistency with the definition of the image centroid of a connected region. An analytical model of the center deviation is derived by comparing the projected centroid with the projection of the true spatial circle center. This deviation is subsequently used to correct the extracted image centroid, thereby improving pose estimation accuracy. Simulation results show that the proposed method significantly enhances the baseline estimation accuracy. Due to the reduced initial errors, further optimization yields limited improvement for yaw and roll angles, while the pitch angle, which is more sensitive to geometric deformation, achieves substantial error reduction (over 55% in most cases). After correction, the overall pose estimation accuracy reaches the arcminute level (approximately 1′), demonstrating the effectiveness of the proposed method for high-precision pose measurement on curved surfaces.