Abstract:Aiming at the difficult problem of aligning the incomplete point cloud obtained from the actual scanning of the laser scanner, a registration method based on corresponding point pairs is proposed. Firstly, experiments were conducted by a laser scanner to obtain the measured point cloud data of the measured workpiece and configuration of Point Cloud Library environment based on Visual Studio software to investigate the point cloud alignment of real and ideal models. Secondly, the measured point cloud data are preprocessed by voxel filtering and uniform downsampling. Then, the point cloud data are aligned by corresponding point pairs to provide a better initial value of the transformation for the subsequent fine alignment. The point cloud alignment is realized based on the ICP algorithm to realize the fine alignment. Finally, the root-mean-square error is used as an evaluation index of the accuracy of the point cloud alignment to evaluate the results of the alignment, and the results are analyzed with the aid of the CloudCompare software. The visual display and analysis show that the root mean square error of the alignment can be controlled at 0.6mm in the case of the surface roughness of the measured workpiece, which proves that the method has better results for the alignment of incomplete point clouds.