并行混合遗传算法在深度像配准中的应用研究

Application research of parallel hybrid genetic algorithm in range images registration

实现了基于并行混合遗传算法的深度像精确配准,并比较了四种不同测度下算法的收敛速度和配准精度。根据进程数将种群划分为相应数量的子种群,每一个进程维护一个子种群的交叉、变异和选择,并通过采用环状的最优个体迁移策略和退火选择算子,实现了基于粗粒度并行混合遗传算法的深度像精确配准。此外,还比较了点对均值、中值、点面距离以及表面间平均体积四种测度下算法的性能和优劣。实验结果表明,并行计算技术的应用能够有效加速遗传算法的收敛,减少算法的运行时间。

The precise alignment of range images is implemented based on a parallel hybrid genetic algorithm.The convergence and alignment results are compared in detail under four different measures.The population of GA is divided into several sub-populations according to the number of processors.Each processor controls the crossover, mutation and selection of each sub-population.One circle topology for elitist migration and annealing selection operator are employed to fulfill the range image registration within the coarse-grained parallel GA.In addition, the performance of four different measures is compared combined with the proposed algorithm.They are mean and median of distance of the corresponding point pairs, and the mean distance of point to the corresponding tangent plane as well as the SMISM measure.Experimental results illustrate that the parallel computing technique can be applied to significantly improve the convergence and reduce the running time of GA.