Abstract
This paper proposed registration of 3D point cloud data sets in reverse engineering of fidelity and precision part by utilization the extracted feature, especially plane features. The objective of proposed method is to solve 3D registration problem of the pairwise data sets with different form and distribution, to improve the fidelity and precision, and to speed up the computation time. This proposed method consists of rough and fine registration. After rough registration, plane features are extracted and validated in correspond grids to be basis of fine registration. Fine registration is then applied based on point-to-point and plane-to-plane registration using ICP by brute force and kD-tree matting scenarios. The experiment showed extraction and validation plane improved the fidelity and precision in both point-to-point and plane-to-plane registration in all scenarios, while kD-tree yielded faster convergence.
