Research And Application Of Grey Wolf Optimization Algorithm

With the continuous advancement of technology and society,solving some complex optimization problems becomes more and more challenging.The traditional solution algorithm has been unable to meet the growing computing needs.In recent years,heuristic algorithms have become one of the effective methods to solve complex optimization problems,and have unique advantages.Grey wolf optimization algorithm is a heuristic algorithm.The algorithm adopts adaptive control parameters and optimal leader diversity strategy.It can find high-quality solutions under multiple constraints,and has the advantages of clear structure and few configuration parameters.However,there are still some problems that need to be solved urgently,such as insufficient theoretical analysis and low accuracy.In order to make up for the shortcomings of the algorithm,firstly,according to the theoretical basis of the grey wolf optimization algorithm,the global convergence analysis of the algorithm is carried out,and then the search mechanism is deeply studied.The grey wolf optimization algorithm combining two improved strategies is used to solve the CEC complex function test set.In addition,the gray wolf optimization algorithm is applied to solve the two-dimensional Otsu image threshold segmentation problem.The main work of this paper is summarized as follows:(1)By introducing martingale theory,the convergence analysis of grey wolf optimization algorithm is simplified.Martingale is a stochastic process,which is widely used to analyze the performance and convergence of stochastic algorithms.Transforming the grey wolf pack evolutionary sequence into a bounded supermartingale can better understand and analyze the convergence properties of the algorithm.By analyzing the boundedness and monotonicity of the grey wolf pack evolution sequence,it can be proved that it converges to a certain random variable in the sense of bounded supermartingale,so as to obtain the probability and speed of algorithm convergence.This method has important application value in algorithm analysis and optimization.It is proved that the grey wolf optimization algorithm has global convergence.This method complements the theoretical basis of the grey wolf optimization algorithm,making the convergence analysis of the grey wolf optimization algorithm more comprehensive.(2)An improved grey wolf optimization algorithm is proposed,which uses the refraction principle of light in physics for reverse learning,and overcomes the problem of low population diversity in the later iteration of grey wolf optimization algorithm.The strategy is not limited to the opposite direction of the individual,while taking into account the other directions of the individual.At the same time,combined with the balanced pool strategy,the global search ability of the algorithm in the later iteration is effectively improved,and the possibility of gray wolf individuals falling into local poles is reduced.The effectiveness of the improved algorithm is verified by 21 commonly used test functions and10 complex CEC 2019 test functions.The experimental results show that compared with other intelligent optimization algorithms and six variants of grey wolf optimization algorithm,the proposed improved algorithm has the characteristics of fast convergence speed,high search accuracy and good stability.(3)Two-dimensional Otsu image threshold segmentation is an important image threshold segmentation method.The gray wolf optimization algorithm combining Circle chaotic mapping mechanism and dynamic position update mechanism is used to optimize the two-dimensional Otsu image threshold segmentation problem.The inter-class dispersion measure function in the two-dimensional Otsu method is used as the fitness objective function.Simple images and complex images are selected for comparative experiments.The comparative experiments show that the gray wolf optimization algorithm combining the two mechanisms has higher segmentation accuracy and better real-time performance than the basic gray wolf optimization algorithm and the classical particle swarm optimization algorithm when optimizing the two-dimensional Otsu image threshold segmentation problem.