动态迁移和椒盐变异融合生物地理学优化算法的高维多阈值分割

针对高维多阈值图像分割中存在的多阈值搜索问题,提出了一种动态迁移和椒盐变异融合的生物地理学优化算法(BBOD)。首先,构建了一种基于动态扰动的迁移算子,对候选解中没有发生迁移操作的特征值添加一个动态的扰动因子,使种群的多样性增加,从而提高全局搜索能力;然后,创建了新型的变异算子,对待变异的特征值产生一个椒盐扰动,使该值在小范围内浮动,以便提高局部搜索能力和算法的收敛速度;最后,将该算法应用到基于最小交叉熵的图像高维多阈值分割中。高维多阈值分割实验结果表明,本文提出的BBOD算法能够获得最优的阈值向量,运行速度、性能指标均优于标准的生物地理学优化(BBO)算法,基于变异的生物地理学优化(BBOM)算法、FFA(Firefly Algorithm)和CSA(Cuckoo Search Algorithm),运行速度是FFA的5倍以上。该算法更适用于基于最小交叉熵的高维多阈值优化选择。

High-dimensional multilevel thresholding based on BBO with dynamic migration and salt & pepper mutation

In view of the threshold search difficulty in high-dimensional multilevel thresholding segmentation, a Biogeography-Based Optimization with Dynamic migration and salt & pepper mutation (BBOD) was proposed. Firstly, a dynamic migration operator was created, and it could add a dynamic disturbance factor to the feature values without migration occured in candidate solutions to increase the diversity of a population. Then,a new type of mutation operator was built to produce a salt and pepper disturbance for the feature values to be mutated,by which the local searching ability and convergence process of the algorithm were accelerated. Finally, the proposed BBOD algorithm was applied to the high-dimensional multilevel image thresholding segmentation based on minimum cross entropy. Experimental results show that BBOD is better in optimization performance and faster in operation speeds than standard BBO (Biogeography-Based Optimization), BBOM(Biogeography-Based Optimization with Mutation),FFA(Firefly Algorithm)and CSA (Cuckoo Search Algorithm),and its operation speed is 5 times as fast as that of FFA. The BBOD is fit to the threshold selection in the high-dimensional multilevel thresholding segmentation based on minimum cross entropy.