Improvement Of Whale Optimization Algorithm And Its Application In Multi-threshold Image Segmentation

Optimization is an engineering mathematics problem with universal applicability.It is difficult for the traditional numerical optimization method to give a reasonable solution in the effective time in the face of large-scale and complex application.However,modern meta-heuristic algorithm can solve these large-scale and complex problems quickly and obtain satisfactory solutions.Whale optimization algorithm is a new global search algorithm proposed by seyedali mirjalili in 2016.This algorithm is one of the most important algorithms in the field of evolutionary computation in recent years,which simulates the hunting mode of humpback whales in the spiral bubble net,and has the advantages of superior mechanism,less parameters involved,simple structure and strong global search ability.However,the whale optimization algorithm still has some shortcomings,such as low stability,slow convergence speed and weak local search ability,which restrict the further application of the algorithm in practice.Aiming at the shortcomings of whale optimization algorithm,this paper makes an in-depth analysis and improvement,aiming to improve the optimization mechanism and performance of whale optimization algorithm,and expand its application in the optimization of function extremum,optimization of engineering design constraints and multi-threshold image segmentation.The main research work of this paper is as follows:(1)A whale optimization algorithm based on piecewise stochastic inertia weight and optimal feedback mechanism was proposed.Firstly,a feedback mechanism based on the current global optimal solution is introduced into the random walk foraging strategy to accelerate the convergence speed of the algorithm and enhance the stability of the solution.At the same time,piecewise stochastic inertia weight is introduced into the contraction enveloping strategy and the spiral bubble net hunting strategy,which not only improves the optimization accuracy,but also enhances the ability of the algorithm to jump out of local extremum.Finally,we modify and improve the transborder processing,which eliminates the potential loss of evolutionary results,and solves the problem of large numbers of individuals crossing the boundary,resulting in large numbers of convergence and loss of diversity.The improved algorithm and other 5 representative algorithms are tested on 12 benchmark test functions and 6 engineering design optimization problems.Thesimulation results show that the improved algorithm is superior to other 5 comparison algorithms in optimization performance,solution stability,applicability and effectiveness for different problems.(2)A whale optimization algorithm with chaotic mapping and iterative local search strategy is proposed.First,Logistic chaotic mapping is introduced into the population initialization stage to increase the population activity and reduce the possibility of the algorithm falling into local extremum.At the same time,the inertia coefficient a and coefficient vector C are improved to adjust the balance between the algorithm's global large-scale exploration and local fine search,so as to accelerate the convergence speed of the algorithm while ensuring the population diversity.Finally,the iterative local search strategy is introduced to further reduce the risk of the algorithm falling into the local extreme value,which is conducive to the algorithm to find the global optimal solution.Theoretical analysis proves that the time complexity of the improved algorithm is consistent with that of the basic whale optimization algorithm and does not reduce the execution efficiency of the algorithm.The experimental results on 13 benchmark functions fully demonstrate that the improved algorithm is superior to other 5 comparison algorithms in optimization accuracy,convergence speed,stability and ability to jump out of local extremum.(3)The whale optimization algorithm based on chaos mapping and iterative local search strategy is applied to solve the multi-threshold image segmentation problem.The coding method suitable for solving the image segmentation problem is constructed,and the fitness function based on the maximum entropy value is defined.The fitness value of the whale's individual location,dimension and objective function is respectively corresponding to the pixel interval,threshold and entropy value of the image.A variety of evaluation methods such as maximum entropy,average execution time,structure similarity,peak signal-to-noise ratio are used to compare and judge the quality of image segmentation.The simulation experiment of segmentation of a large number of images by seven contrast algorithms shows that the improved algorithm is an efficient and practical image segmentation method,which can quickly segment images,accurately select the threshold and improve the image segmentation quality.