Fusing multi-source data and point clouds to reconstruct ancient architecture

Ancient architecture carries rich cultural information, but its structure is complex, and its texture details are numerous. Traditional three-dimensional (3-D) reconstruction methods are difficult to fully capture its information. To achieve high-precision 3-D reconstruction of ancient buildings, an ancient architecture reconstruction on the basis of multi-source data fusion and a point cloud registration algorithm is constructed. By fusing multi-source data and combining voxel quotient filtering, curvature feature point extraction, and the Poisson reconstruction algorithm, ancient buildings are finely reconstructed. Then, the point cloud registration algorithm is introduced to optimise the accuracy and efficiency of data alignment. The collaborative fusion framework for ancient architectural data based on a multi-source data fusion model could achieve data fusion accuracy and model integrity of 96.5% and 98.1%, respectively, in the ancient architecture reconstruction. When the number of point clouds is 500 000, the model registration time is 114 s, and the model accuracy is 0.5 mm. When the number of point clouds is 1.2 million, the model registration time is 152 s, and the model accuracy is 0.6 mm. The registration time and model accuracy are better than other reconstruction models. The research provides more efficient and accurate methods for cultural heritage protection, which has important practical value.