面向Otsu阈值搜索的PSO惯性因子改进方法

Modified strategy to inertia weight in PSO for searching threshold of Otsu rule

针对搜索最大类间方差法(Otsu)的最优阈值,提出了一种粒子群优化(PSO)算法中惯性因子的改进方法。该方法使粒子群优化算法中的惯性因子与粒子群的群最优适应值和个体最优适应值相关,用个体最优适应值与群最优适应值之间的距离修正惯性因子的大小。随着个体最优适应值与群最优值之间距离的减小,惯性因子也相应减小,反之则增加。实验结果表明:该方法与已有PSO惯性因子的进化方法相比,计算量小,收敛速度平均提高了21.072 6%。使用该方法可改善图像在线分割速度。

In order to searching the best threshold of Otsu rule rapidly, a modified strategy to the inertia weight in Particle Swarm Optimization （PSO） is introduced. The strategy correlates the inertia weight in PSO with the globally best position and the individual best position in particle swarm and modifies the inertial weight by the distance between the best individual position and the globally best position. With the reduction of the distance between the globally best position and the individual best position, the inertia weight of the individual particle will be reduced, vice versa. The experimental re sults indicate that this algorithm decreases computational work and improves average convergence rate by 21. 072 6% as compared with that of traditional methods. It can speed up on-line image segmenta tion using this method.