Abstract: Unconventional reservoirs such as shale oil exhibit pronounced anisotropy resulting from the preferential alignment of clay minerals and micro-pores, giving rise to highly complex seismic responses. Conventional inversion methods often fail to accurately characterize anisotropic parameters, thereby limiting reliable fracture characterization and comprehensive reservoir evaluation. To achieve efficient and high-precision reservoir characterization, this study replaces the isotropic component in the Rüger approximation with the exact Zoeppritz equations. This modification improves the accuracy of reflection coefficient calculations at wide incidence angles and enhances the sensitivity of the inversion formulation to anisotropic parameters. In addition, a Kalman localization strategy is introduced to effectively mitigate spurious correlations associated with limited ensemble sizes. Based on these improvements, a comprehensive inversion workflow is developed, including localization-scale analysis, prior geostatistical simulation, multi-parameter iterative updating, and uncertainty quantification. Both synthetic experiments and field-data applications demonstrate that the proposed method can achieve stable and high-accuracy inversion of elastic and anisotropic parameters, even with relatively small prior ensembles. The inversion results show strong consistency with geological features and well-log observations. By effectively balancing computational efficiency and robustness against noise, the proposed method provides reliable geophysical support for subsequent fine-scale fracture characterization and sweet-spot evaluation in shale oil reservoirs.