3D registration process of ICP using linear constraint

Authors

  • Zihuan Ding

DOI:

https://doi.org/10.56028/aetr.4.1.542.2023

Keywords:

Iterative closest point (ICP),3D registration, linear transformation, sub- sampling, point-to-point constraint, point-to-plane constrai.

Abstract

This paper implements an ICP registration algorithm that aligns mul- tiple 3D scans into a common coordinate system. To better visualize the result, we calculate the vertex valence using a triangle mesh and visualize the 3D object in color based on the vertex valence. The initial job of im- plementing the ICP algorithm is to evenly sub-sample the 3D scans to in- crease the calculation speed, compute the correspondence, and remove the bad corresponding pairs based on distance threshold and normal compat- ibility. Subsequently, perform the registration process with point-to-point and point-to-surface registration, which transforms the ICP equation by solving a matrix equation with a linear constraint. The result shows that both methods recreate a complete 3D object, but the point-to-surface method converges much faster.

Downloads

Published

2023-03-18