Accurate Registration Technology of Ancient Architectural Point Cloud Based on Improved ICP Algorithm and Its Application in Digital Protection

This article proposes an accurate registration method for ancient architectural point clouds based on an improved ICP algorithm, and applies it to the digital protection of architectural heritage. With the rapid development of 3D LiDAR scanning and unmanned aerial vehicle photogrammetry technology, the application of 3D point cloud data and image data in virtual restoration of historical buildings is becoming increasingly widespread. This article comprehensively utilizes drone oblique photography and ground LiDAR scanning technology to obtain complete historical building point cloud data, and based on SVD’s Scale ICP algorithm, achieves high-precision registration of image dense point cloud and ground LiDAR point cloud. To address the issue of low accuracy in edge detection during point cloud completion, this paper proposes an improved PCA vector estimation method, which enhances the accuracy and noise resistance of edge detection. In addition, this article proposes a context based point cloud completion strategy that improves the efficiency and accuracy of point cloud completion by first segmenting and then iteratively completing based on blocks. The experimental results using historical buildings in Gan County as the research object show that the method proposed in this paper can effectively improve the accuracy and completeness of the point cloud model of ancient buildings, providing strong support for digital protection technology of ancient buildings.