Effective Multiview Registration of Point Clouds Based on Student’s-t Mixture Model

Yanlin Ma,Jihua Zhu,Zhiqiang Tian,Zhongyu Li

DOI: https://doi.org/10.1016/j.ins.2022.06.006

IF: 8.1

2022-01-01

Information Sciences

Abstract:Recently, the expectation maximization (EM) algorithm has been used to address the multiview registration of point clouds problem. Most approaches suppose that each data point is generated from the Gaussian mixture model, which notably has difficulty handling outliers and heavy-tail noise. Consequently, this study proposes a novel multiview registration approach based on Student’s-t mixture model (SMM). Specifically, we assume that each data point is generated from a unique SMM, where its nearest neighbors (NNs) in other point clouds are SMM centroids with fixed degrees of freedom, equal covariances, and membership probabilities. Therefore, the multiview registration problem is formulated as the maximization of the likelihood function. Subsequently, the EM algorithm is employed to optimize the rigid transformations used for multiview registration and the only Student’s t-distribution covariance. Because only a few model parameters are optimized, our algorithm achieves promising registration performance. Additionally, all SMM centroids are obtained using the NN search algorithm, which is exceedingly efficient. Moreover, the Student’s t-distribution renders our algorithm inherently robust to outliers and heavy-tail noise. The results of tests on benchmark datasets demonstrate our algorithm’s superior registration performance in terms of accuracy and robustness, compared with state-of-the-art algorithms.