Benchmark Based Selection of Point Cloud Registration Algorithms for Cultural Heritage 3D Digitization

This work aims to address the point cloud global registration algorithm selection problem by proposing a pipeline for testing and evaluating such algorithms. Moreover, this benchmark-based evaluation automates the algorithm selection process by assessing the point cloud registration algorithms under the same universal benchmark, using complete and diverse datasets. We demonstrate its performance by comparing two State-of-the-art global registration algorithms that are tested on a set of problems created by diverse, unstructured datasets, acquired by various types of sensors, and in different environmental conditions. Furthermore, a comprehensive metric is proposed, which serves as a qualitative metric, rather than an absolute result for the evaluation process. Finally, the different outcomes of the algorithms benchmarking performance are thoroughly presented and discussed.