Application Of Improved Butterfly Optimization Algorithm In Image Segmentation

As a new meta heuristic algorithm,butterfly optimization algorithm(BOA)has attracted the attention of researchers.It has good competitiveness in solving numerical optimization problems and practical applications.The algorithm can simulate the feeding process and mating situation of butterflies.Because the parameters of BOA algorithm are less,only olfactory perception rules are considered,and the iterative process is simple and easy to implement.However,the convergence rate is slow,and it is easy to fall into the local optimum.Therefore,the butterfly optimization algorithm needs to be improved to increase population diversity and balance the global and local exploration capabilities.To solve the above problems,a butterfly optimization algorithm with memory function and adaptive weight(MBOA)and a butterfly algorithm with scheduling processing and dynamic threshold adjustment(DBOA)are proposed respectively.At the same time,the improved algorithm is applied to image segmentation,achieving a high research goal.The main work of this thesis is divided into three parts:First,a butterfly optimization algorithm(MBOA)with memory function and adaptive weight is proposed to solve the problems of slow convergence rate and poor population diversity of butterfly optimization algorithm.First,chaos map is used in population initialization to control the initial distribution of butterflies.Secondly,two random factors and learning factors are introduced in the local search stage,so that the butterfly can learn from the individual’s historical optimal position and avoid falling into the local optimal solution region.Thirdly,Cauchy mutation is used to increase the diversity of the population and improve the global search ability of the algorithm.Finally,a dynamic weight value is designed to effectively balance the exploration ability of the algorithm and ensure that butterfly individuals can better search in the middle group space.The experimental results show that the convergence speed and precision of MBOA are improved.Second,a butterfly algorithm(DBOA)with scheduling and dynamic threshold adjustment is proposed to solve the problem of BOA using fixed threshold in the search process.First,elite reverse learning is applied to the initialization stage of BOA.Secondly,dynamic adjustment of probability threshold is introduced to balance the proportion of global exploration and local search.Thirdly,the butterfly is scheduled to search near the optimal solution to improve the accuracy of optimization.Finally,according to the fitness of butterfly individuals,the variance is calculated to determine whether to search in a small range around the target.The experimental results show that DBOA has better improvement effect in optimization rate and convergence precision than other algorithms.Third,in the medical field,image segmentation has been widely used as a general image processing method.In order to improve the texture clarity of the segmented image,this thesis adopts the two-dimensional Renyi gray entropy multi threshold segmentation method,and applies the improved butterfly optimization algorithm to the image segmentation,which solves the problem of missing key information,messy texture and most shadows of the segmented image.The experimental results show that MBOA algorithm and DBOA algorithm can greatly improve the sharpness and contour quality of the segmented image.