Precise Iterative Closest Point Algrithm Based on Correntropy for 3-D Oral Data Registration

This paper proposes a new iterative closest point approach based on correntropy with feature guided. Iterative Closest Point (ICP) algorithm can deal with most rigid registration problems, but for point sets with lots of noise and outliers, ICP cannot achieve high precision. We introduce correntropy into ICP to handle this problem by suppressing the influence of the noise and outliers. In terms of point sets contain a large proportion of planes or a curved surface, and have single structure, such as a three-dimensional model of upper jaw, we propose a feature-guided model to solve the oral data registration problem, which uses both the feature and the original data to participate in the registration, but with different weights. Our method mainly deals with the point set registration which has single structure and contains outliers. Experimental results demonstrate that the proposed algorithm is precise and robust.