As an important component of cultural heritage,the refined protection of ancient architecture urgently requires high-precision 3D data support.However,problems such as noise interference and multi-source data registration errors during point cloud acquisition lead to insufficient 3D reconstruction accuracy, which seriously restricts the process of digital protection.To achieve precise registration of point cloud data of ancient buildings,a collaborative optimization framework of multi-source mixed filtering and coarse fine registration algorithm is proposed.For the heterogeneous data of unmanned aerial vehicle (UAV) aerial survey and ground liDAR,octree voxel grids are used for decimation.Combined with fabric simulation filtering, statistical filtering and pass-through filtering,noise points and outliers are removed to retain the detailed features and structural information of ancient buildings. An improved principal component analysis method is used to construct a principal component orthogonal reference frame, the initial pose is optimized through principal axis alignment,and the iterative nearest point method is combined to achieve point cloud registration.The results show that the number of point clouds decreased to 15.15%and 91.36% after hybrid filtering.The accuracy and recall of the registration method on the Model Net40 dataset reach 98.12% and 97.55%,respectively,which is 6.31% higher than traditional PCA.The root mean square error of large-scale point cloud registration is optimized to 0.1954m,and the registration time is only 882.56 seconds.The research significantly improves the processing efficiency of point cloud data of ancient buildings through the collaborative mechanism of multi-source filtering denoising and high-precision registration, providing a feasible technical path for the digital protection and 3D reconstruction of cultural heritage.