Information granule-based multi-view point sets registration using fuzzy c-means clustering

This paper addresses the registration problem for multi-view point sets. Motivated by the formation of information granule and casting registration as a clustering task, an information granule-based multi-view point sets registration using fuzzy c-means clustering is proposed. Information granules are formed following the principle of justifiable granularity, and the data points covered by information granules can be obtained to represent the structural crux of the point set. The preprocessing step using information granule can achieve point set simplification and enhance the robustness of subsequent registration. Then, the aligned point sets involved in multi-view registration are clustered, and fuzzy clustering is used to solve the clustering problem and multi-view registration problem simultaneously. Membership function is introduced into the clustering-based registration, which improves the registration performance in comparison with other clustering-based methods with hard partition. Finally, the clustering and transformation estimation are alternately and iteratively applied to all point sets until the final clustering and registration results are obtained. Experiments using publicly benchmark datasets demonstrate that the proposed approach achieves better performance than the comparison approaches in terms of the accuracy and robustness for multi-view registration.