针对传统狼群算法(WPA)应用于图像分割寻优慢且易陷入局部最优的缺点, 提出一种基于分数阶狼群的Otsu图像分割算法。利用分数阶微分对过去状态有记忆性的优点, 用分数阶阶次来控制狼群在游走过程中的位置更新, 并引入自适应分数阶阶次, 根据狼的位置信息自适应地调整分数阶阶次, 提升算法收敛速度。采用粒子对称分布方法改进狼群围捕行为, 改善狼群个体空间分布状态, 提高种群多样性, 调整围捕过程中的狼群位置, 克服算法后期易出现局部最优的弊端。采用改进的二维Otsu算法, 将其离散度矩阵作为狼群算法的寻优函数,将目标从图像中分割出来。实验表明, 本文算法达到稳定的收敛次数较传统狼群算法平均提升了50次左右, 较文献<参考文献原文>提出的算法平均加快了10次左右。本文改进算法保证了图像分割精度, 并提升算法收敛速度。

The traditional wolf pack algorithm (WPA) is slow and easy to fall into local optima in the image segmentation. Aiming to this problem, the Otsu image threshold segmentation algorithm based on fractional order wolf pack algorithm is proposed. Using the advantage of fractional order which has memory for past states, the position updating of wolves is controlled by fractional order. Using the adaptive fractional order, the position of wolves is used to adjust the fractional order adaptively to improve the convergent speed. The particle symmetry distribution method is used to improve the hunting behavior, improve the spatial distribution of wolves, adjust the position of wolves during hunting, overcome the shortcomings of the algorithm. Using the gray-gradient two-dimensional histogram, and inter-class variance of Otsu algorithm defined as the fitness function, the target is segmented from the image. Experimental results show that the number of convergence times achieved by this algorithm is about 50 times higher than that of the WPA, and about 10 times faster than the algorithm proposed in literature <参考文献原文>. The algorithm in this article can not only accelerate the convergent speed, but also ensure the accuracy of image segmentation.