Jiandu point cloud registration using high-resolution data and generalized t-student kernel

As a crucial carrier of Chinese civilization, Jiandu artifacts present significant challenges for high-precision point cloud registration in digital restoration of fragmented pieces. To address limitations of existing methods in handling complex fracture surfaces, noise sensitivity, and incomplete data, this study proposes a registration framework integrating RGB-D data, local geometric features, and a Generalized T-Student kernel. A correction algorithm based on local normals and regional connectivity enhances fracture surface representation through multi-scale geometric descriptors. A texture gradient direction consistency rule is introduced, embedding semantic texture constraints into the ICP process to ensure both geometric alignment and texture continuity. Additionally, a robust optimization model with adaptive weighting suppresses noise and outliers. Ablation and simulation results demonstrate that the proposed method outperforms classical and state-of-the-art approaches in precision and robustness, supporting more accurate digital preservation of cultural heritage.