摘要
This study reduced mid-spatial-frequency (MSF) errors in optical component polishing by using subaperture techniques such as bonnet polishing. MSF errors can degrade optical imaging quality by causing light scattering and reducing system resolution. By optimizing a spiral path and incorporating machine learning techniques, such as support vector regression (SVR) and Gaussian process regression (GPR), the SVR and GPR models achieved an R2 value exceeding 85%. The optimized root mean square error (RMSE) of the power spectral density (PSD) value was 0.93 × 10 −3 µm ∧ 3, with the maximum and minimum error reductions being 97.0% and 10.4%, respectively. Thus, the optical surface quality improved substantially, indicating the effectiveness of optimization techniques in reducing the MSF error.