Research On Multi-threshold Image Segmentation Based On Multi-strategy Fusion Improved Swarm Intelligent Optimization

Image segmentation is a very key part of digital image processing technology,which divides the image into non-overlapping parts with different characteristics.In practical engineering application,it is very important to extract and analyze the result of image segmentation for image recognition and other fields.Multi-threshold image segmentation is a kind of method with good segmentation effect in image segmentation technology,but with the increase of the number of selected thresholds,the computational complexity of image segmentation process increases exponentially.Therefore,a more efficient and stable multi-threshold segmentation method is very necessary.Swarm intelligent optimization algorithm is a kind of optimization algorithm which attracts more attention among intelligent optimization algorithms.It provides an effective method to obtain the optimal solution.This paper aims to apply the improved swarm intelligent optimization algorithm to obtain a group of optimal thresholds in the multi-threshold image segmentation,so as to improve the accuracy and efficiency of image segmentation.Based on the studied algorithm,an improved multi-threshold segmentation method is proposed to effectively segment wood defect images.(1)Aiming at the common problems of population diversity loss and slow convergence of swarm intelligent optimization algorithm,chaotic population initialization strategy,elite opposition-based learning strategy and chaotic elite opposition-based learning strategy are introduced to optimize swarm intelligent optimization algorithm.In order to select the method with the best optimization effect among the three initialization strategies,The three strategies are combined with chimp optimization algorithm,northern goshawk optimization algorithm and eagle perching optimizer algorithm respectively,and the performance of the algorithm was tested on the benchmark function.Experiments show that the improved swarm intelligent optimization algorithm based on Gaussian chaos elite opposition-based learning strategy has the most outstanding global search ability and the fastest convergence speed in most scenarios.(2)Aiming at the problem of weak local search ability and easy to fall into local extremum of swarm intelligent optimization algorithm,Gauss Cauchy mutation strategy,dimension-bydimension barycenter reverse mutation strategy and hybrid opposition-based learning strategy are introduced to optimize the swarm intelligent optimization algorithm..Based on the improved swarm intelligent optimization algorithm based on Gaussian chaotic elite opposition-based learning strategy,experiments are carried out to compare the anti-stagnation ability of three optimized individual iterative update strategies.Experiments show that the convergence accuracy and efficiency of the swarm intelligent optimization algorithm based on the Gauss chaotic elite opposition-based learning strategy and Gauss Cauchy mutation strategy are the most prominent.(3)Aiming at the problems of high computational complexity and low image quality after segmentation in traditional multi-threshold image segmentation,a multi-threshold image segmentation method based on multi-strategy fusion swarm intelligent optimization algorithm is proposed.Firstly,on the basis of the first two chapters,the multi-strategy fusion swarm intelligence optimization algorithm with the strongest optimization ability is selected and combined with three multi-threshold image segmentation techniques: Otsu,Kapur entropy and symmetric cross entropy.The segmentation results of three segmentation methods for Berkeley color images are compared and analyzed,and the multi-threshold segmentation method with the best comprehensive index is selected.Secondly,in order to further verify the superiority of the proposed segmentation method,the classical swarm intelligence optimization algorithm and the multi-threshold segmentation method optimized by the improved swarm intelligence optimization algorithm are selected as the comparison algorithm,and the wood defect image segmentation experiment with more complex experimental background is carried out again.