Digital Restoration of Fragmented Ancient Pottery Using Improved Contour Extraction and Point Cloud Registration

To address the challenges of low accuracy and high computational cost in the virtual reassembly of ancient pottery fragments—caused by irregular edge shapes, sparse fracture features, and large-scale data processing—this study proposes a novel method based on an adaptive neighborhood density α-shape algorithm for contour extraction. By incorporating an alpha determination strategy and multi-resolution analysis, the algorithm accurately identifies contour triangles and extracts precise fragment edges. On this basis, a two-stage registration approach is adopted: coarse alignment followed by fine registration using an improved ICP algorithm enhanced with pseudo-corresponding points, thereby determining the optimal spatial transformation between fragments. Experiments were conducted using pottery fragments unearthed from the Houzhai Cemetery in Suozhou, Shanxi Province. The reassembly success rates reached 87% for a set of 39 fragments and 90% for a set of 30 fragments, with overall computational efficiency improved by 15%–20% compared to existing methods. The results demonstrate that the proposed method offers an effective technical solution for the digital restoration of fragmented cultural relics.