No abstract

Abstract. 3D registration is an essential step in 3D reconstruction of real objects. In this paper, the fine registration methods for point clouds are classified systematically and analyzed theoretically. For several typical fine registration algorithms, experimental comparisons are also carried out on the effect of sampling parameters, iterations, iterative threshold and random noise. Furthermore, the stability and efficiency of the algorithms are tested. Based on the experimental results, the problems in current registration methods and the prospects of future research directions are also presented.