ICP registration using invariant features

Abstract

This paper investigates the use of Euclidean invariant features in a generalization of iterative closest point registration of range images. Pointwise correspondences are chosen as the closest point with respect to a weighted linear combination of positional and feature distances. It is shown that under ideal noise-free conditions, correspondences formed using this distance function are correct more often than correspondences formed using the positional distance alone. In addition, monotonic convergence to at least a local minimum is shown to hold for this method. When noise is present, a method that automatically sets the optimal relative contribution of features and positions is described. This method trades o error in feature values due to noise against error in positions due to misalignment. Experimental results show that using invariant features decreases the probability of being trapped in a local minimum, and is most e ective for di cult registration problems where the scene is very small compared to the model.

Coarse registration of 3D surface triangulations based on moment invariants with applications to object alignment and identification

“…Lots of the extensions to this now standard 3D registration method are compared in [16]. The application of moment invariants from 3D points to the ICP algorithm is shown in [18]. Compared to our work, in [18] the usage of moment invariants is incorporated directly into the iterative estimation procedure of ICP, which still needs an initial solution.…”

Higher Dimensional Affine Registration and Vision Applications

“…The ICP algorithm is a very general point registration algorithm that has become popular in vision applications, e.g., [1], [12], [15], [21], [37], [44], [57]. While these works are concerned exclusively with rigid transformations in IR 2 ; IR 3 , it is not difficult to extend the ICP algorithm to incorporate affine transformations.…”

Pair-wise Registration of 3D/Color Data Sets with ICP

“…Rusinkiewitcz & Levoy [12] have classified most of these variants and evaluated their effect on the algorithm convergence. According to the Rusinkiewitcz classification, we consider that the variants proposed for ICP concern the different steps of the algorithm: selection of points to be registered [8,10,12], matching technique [4,6,7,11,13], weighting of the matched pairs [7,9], outliers rejection [10,11,15] and transformation estimation [1,2,4,5,9,14].…”