Sun meridian extraction from skylight polarization patterns via mirror symmetry and jump-line fusion

Accurate extraction of the solar meridian from skylight polarization patterns remains challenging under variable atmospheric conditions. To address this limitation, we develop a computational framework that synergistically integrates two fundamental characteristics of Rayleigh scattering: the mirror symmetry in degree of linear polarization (DOLP) distributions and the antisymmetric discontinuity in angle of linear polarization (AOLP) profiles. Our approach begins with robust polarization field preprocessing through interquartile range (IQR)-based outlier rejection, followed by dual-path analysis in polar coordinates space. The mirror symmetry axis is optimized via mean square error (MSE) minimization constrained by neutral point positions, whereas AOLP jump-lines are detected through complex-domain transformation combined with radial-angular segmentation. These complementary estimations are then adaptively fused using confidence weights derived from their respective reliability metrics. Based on extensive experimental validation, the proposed fusion method achieves a high degree of alignment with the astronomical solar azimuth, with an average MSE of 0.02 (sunny), 0.051 (overcast), 0.053 (fog), and 0.017 (dust) across diverse atmospheric conditions. The orientation estimates are further stabilized via an adaptive temporal filter, effectively suppressing outliers by enforcing the physical continuity of solar trajectory. The proposed method not only provides a reliable polarization-based heading reference for autonomous navigation systems but also advances computational methodologies for atmospheric polarization analysis.