An Adaptive Differential Evolution Algorithm with a Point-Based Approach for 3D Point Cloud Registration

This paper presents a novel pairwise registration approach, which aligns images of the same object that have different ranges. By using a point search medium instead of a conventional six-dimensional parameter to reduce the number of search dimensions, the new method resulted in a higher convergence rate and robustness in the same search conditions. The approach integrated a hybrid registration strategy, a combination of Iterative Closest Point (ICP) as a local aligning tool and a global search algorithm such as simulated annealing, particle swarm optimization, differential evolution, etc. An adaptive differential evolution algorithm called ISADE was chosen as the best-so-far global search algorithm. Different experiments on different datasets were carried out. In the new method, as compared with the conventional approach, better aligning results in convergence rate and robustness were observed.