A Point Cloud Stitching Approach Based on Image Registration for High-Precision Threaded Surface Modeling in Multi-View 3D Imaging

With the rapid development of 3D imaging technology, point cloud models have found wide applications in various fields, such as 3D building modeling and precision manufacturing quality inspection. In the oil and gas industry, the threaded connections of casing pipes, as critical components, have complex surface geometries and require high precision. Due to the limitations of a single-view scan window, it is impossible to capture the complete geometric features of the threaded surface, making it necessary to register point cloud data from different viewpoints to achieve accurate alignment and generate a complete 3D model. This paper proposes a point cloud stitching method using image registration. The SIFT algorithm extracts and matches feature points from adjacent grayscale images, while RANSAC optimizes the matching for accuracy. A weighted averaging algorithm fuses depth maps, which are then converted into 3D point clouds based on the camera's intrinsic parameters. Experimental results show that this method outperforms traditional point cloud registration methods in both accuracy and efficiency, enabling fast and precise stitching of multi-view point cloud data for thread parameter extraction and quality inspection.