PROJECT TITLE :

Joint Alignment of Multiple Point Sets with Batch and Incremental Expectation- Maximization - 2017

ABSTRACT:

This paper addresses the problem of registering multiple purpose sets. Solutions to the present problem are typically approximated by repeatedly solving for pairwise registration, that results in an uneven treatment of the sets forming a combine: a model set and a data set. The main disadvantage of this strategy is that the model set might contain noise and outliers, which negatively affects the estimation of the registration parameters. In contrast, the proposed formulation treats all the point sets on an equal footing. Indeed, all the points are drawn from a central Gaussian mixture, hence the registration is forged into a Clustering drawback. We have a tendency to formally derive batch and incremental EM algorithms that robustly estimate each the GMM parameters and the rotations and translations that optimally align the sets. Moreover, the mixture's means play the role of the registered set of points while the variances give rich data about the contribution of every component to the alignment. We tend to thoroughly check the proposed algorithms on simulated information and on challenging real data collected with range sensors. We compare them with several state-of-the-art algorithms, and we tend to show their potential for surface reconstruction from depth data.