Point cloud optimization of multi-view images in digital image correlation system

Post-processing of the initial point cloud is one of the important ways to improve the measurement accuracy of digital image correlation (DIC). In this study, a multi-level point cloud optimization scheme is proposed for processing point clouds of multi-view images in DIC systems. Dense point clouds obtained after interpolation of primary point clouds from multi-view images contain noise. First, the local Random Sample Consensus (RAN -SAC) algorithm is proposed to retain local features while removing coarse matching points. Defects are filled in by the adaptive cavity repair technique based on information around the cavities and restores the original surface shape to the greatest extent. Finally, a multi-layer surface filtering is executed to smooth the mesh. Experiments verify the comprehensive performance of the multi-level optimization scheme, providing excellent results in point cloud fusion, cavity repair, and mesh smoothing, regardless of static 3D reconstruction or dynamic deformation measurement.