A Process for Non-Rigid Fine Object Registration Based on Open3D

The massive demand for three-dimensional scanning of the physical world and artificial modeling makes three-dimensional reconstruction more important. This article is aimed at the three-dimensional reconstruction of the human foot, due to the displacement and the possible deformation of the foot in the scan (such as surface shrinkage and relaxation caused by respiration) and the systematic error of the scanner, the point cloud needs to introduce rigid registration and non-rigid registration to improve the accuracy of subsequent mesh reconstruction. The main tasks include: providing a point cloud stitching algorithm that calculates the overlapping parts of adjacent scan frames and then uses non-rigid registration; practically proposing a set of fine small pieces of rigid and non-rigid objects 3D reconstruction of the system scheme. Experimental results show that point cloud stitching using non-rigid registration improves the accuracy of modeling compared with direct point cloud stitching.