基于分数阶粒子群的Otsu图像分割算法

摘要

To avoid the traditional particle swarm optimization algorithm falling into local optima,the Otsu image thresholding segmentation (ImFpsoOtsu) algorithm was proposed based on fractional order particle swarm optimization.The grayscale-gradient two-dimensional histogram was used,and inter-class variance of Otsu algorithm was defined as the fitness function.The particle evolution factor was introduced and the particle state information was used to adjust the fractional α,and the speed and the position were updated by setting speed to zero.An improved particle swarm optimization algorithm combined with the traditional particle was used to update the formula and the particle symmetry distribution was used to obtain the optimal threshold,and the target was segmented from the image.Experimental results show that the proposed algorithm can ensure the image segmentation effect and improve the convergence speed of the algorithm effectively.%为避免传统粒子群优化算法陷入局部最优,提出基于分数阶粒子群优化的Otsu图像阈值分割(ImFpsoOtsu)算法.采用基于灰度级-梯度二维直方图算法,以Otsu算法的最大类间方差为适应度函数;通过引入粒子进化因子,利用粒子的状态信息自适应更改分数阶次α,通过速度增量为零来更新粒子速度、位置值;结合传统粒子群粒子更新公式,采用粒子对称分布的改进粒子群算法获取最佳阈值,将目标从图像中分割出来.实验结果表明,所提算法保证了图像的分割效果,有效提升了算法的收敛速度.