A Global Feature-Less Scan Registration Strategy Based On Spherical Entropy Images

Scan registration is the fundamental of several advanced 3D data processing techniques, and the majority of existing global registration methods depend on explicit features. This paper presents a global feature-less scan registration strategy based on the Spherical Entropy Image (SEI) and the Generalized Convolution Theorem. The structure of the scan is described by a spherical function named SEI in this paper. The 3D rotation is then estimated by aligning the corresponding SEIs. After that, the Phase Only Matched Filtering (POMF) is adopted for translation recovery. No particular features in the input data are prerequisite to our method. Unlike the feature-based methods, the performance of our method does not reply on specific proper parameters. The algorithm is validated using the challenging data captured by our custom-built platform and publicly available datasets. The experimental results illustrate the parameter-independence, high reliability and efficiency of our novel algorithm in registration of feature-less scans.