SCCA–HRBF: A Structure Confidence-coupled Adaptive Hermite Radial Basis Function Method

Abstract

 A structure confidence-coupled adaptive Hermite radial basis function (SCCA–HRBF) method is proposed for denoising precision optical measurement point clouds affected by random noise, local anomalous disturbances, and varying observation reliability. First, local curvature features are extracted using principal component analysis of neighboring regions, and an observation confidence model is constructed by integrating height and normal consistencies and local fluctuation information to evaluate the reliability of measured points. Subsequently, a structure confidence-coupling mechanism is established to jointly characterize local structural complexity and observation reliability, enabling adaptive prioritization of HRBF centers and nonuniform degree allocation for implicit surface reconstruction. Finally, point-value, gradient, and local structural consistency constraints are integrated within a weighted Hermite radial basis function framework to achieve continuous surface reconstruction, followed by point-cloud correction along the measurement direction. The proposed method is validated using real point-cloud data acquired from mosaic detector focal planes and compared with Gaussian, bilateral, moving least-squares (MLS), and conventional HRBF methods. The results demonstrate the superior performance of SCCA-HRBF in terms of the root-mean-squared error (RMSE), mean absolute error (MAE), Peak-to-Valley (PV), Maximum Error (MAXE), and standard deviation (STD). Specifically, the RMSE and RV are reduced to 0.300688 μm and 1.186177 μm, respectively. Center-number and parameter sensitivity analyses verify the effectiveness of the proposed nonuniform degree allocation strategy and the method’s robustness. These results indicate that the SCCA–HRBF effectively suppresses abnormal measurement disturbances and improves the accuracy and robustness of optical surface reconstruction, providing an effective data-processing approach for mosaic detector focal-plane inspection and precision optical surface measurements.