Point Cloud Registration Based on Improved Tuna Swarm Optimization Algorithm

To enhance the performance and efficacy of point cloud registration, this research paper proposes a registration methodology centered on an enhanced Tuna Swordfish Optimization (TSO) algorithm. In order to tackle the issue of non-uniform population distribution resulting from random initialization in TSO, the Tent chaotic mapping technique is employed to initialize the population parameters, thereby augmenting population diversity. To prevent convergence to local optima, the Levy flight strategy is introduced to fortify the system's escape capability. Through the utilization of a nonlinear weight factor, the algorithm effectively balances global exploration and local optimization abilities. Ultimately, the fine registration is achieved by employing the Iterative Closest Point (ICP) algorithm. Experimental results, conducted on the Bunny model of a publicly available dataset, demonstrate that the algorithm significantly enhances alignment accuracy by 66.73% and 25.73% when compared to the ICP algorithm and the ITSO algorithm, respectively. Furthermore, the algorithm's validity is corroborated through its application to multiple datasets.