Improvement Of Firefly Algorithm And It Application In Image Segmentation

Firefly algorithm is an optimization technology based on swarm intelligence.It has the characteristics of high parallelism,self-organization,self-learning and self-adaptation,and has been proved to have good performance in solving various optimization problems.However,firefly algorithm also has the common faults and defects of swarm intelligence algorithm,such as parameter selection,premature convergence,easy to fall into local optimal,weak theoretical basis,etc.In this thesis,on the basis of the study of firefly algorithm principle,based on the mathematic based on the theoretical analysis and numerical performance verification,from the parameter adaptive,structure optimization,the model optimization Angle three improved strategy is proposed and a model,and combined to reconstruct a kind of adaptive algorithm,the proposed improved algorithm in the benchmark test function experiment,the results show that the These strategies can effectively improve the accuracy of the solution and reduce the computational time complexity.In addition,the image segmentation problem as a classical problem in the field of computer vision,is one of the most difficult problems in image processing,is also a kind of complex optimization problem,promote the firefly algorithm to solve practical problems.This thesis improved the firefly algorithm was applied to image segmentation problem,with efficient optimization method to realize the automatic image processing and analysis technology.The results of the above methods on the Berkeley dataset also show the effectiveness and robustness of the improved firefly algorithm for image segmentation.The main research work of this thesis is summarized as follows:Three improved strategies and an attraction model are proposed,and an adaptive firefly algorithm based on covariance elite selection is reconstructed.Aiming at the problem that the step size of the standard firefly algorithm may fall into local optimum in the iterative process,leading to premature convergence,an adaptive step size strategy is proposed,which can dynamically adjust the step size and decrease exponentially.In order to retain several individuals with the best fitness and most likely to approach the optimal solution,a strategy of elite selection based on covariance matrix was proposed.Aiming at the problem that the neighborhood search operator is stuck in local optimum,a "punishment" mechanism is proposed to destroy the structure of local optimum solution and make the operator jump out of local optimum.Aiming at the problem of too many times of attraction in the total attraction model of standard firefly algorithm,a bidirectional guidance model is proposed.The main idea of this model is that fireflies will be affected by an elite and a random individual in a population at the same time in a movement,and the time complexity of the algorithm is greatly reduced due to the reduction of the number of attracting.Finally,according to the characteristics of each stage of the algorithm,three improved strategies and one attractor model are combined and reconstructed,and the performance of the improved algorithm is significantly improved,which has strong adaptability and balances the local mining and global exploration ability of the algorithm well.Apply the improved firefly algorithm to the image segmentation problem.In the traditional gray threshold segmentation problem,Kapur entropy and Otsu interclass variance method are used as objective functions,and the improved firefly algorithm is used to find the optimal threshold.For color image segmentation,the proposed method is divided into three consecutive stages.In the first stage,a Gaussian filter is used to smooth the three-dimensional histogram based on RGB color space to eliminate the unreliable and non-dominant peak values in the histogram that are too close to each other.In the next stage,the improved firefly algorithm is used to identify the peaks representing different clusters in the histogram.Finally,pixel clustering is performed according to Euclidean distance,and the number of clusters can be determined automatically without manual pre-definition.