Research On Image Threshold Segmentation Method Based On Swarm Intelligence Algorithm

Image segmentation,as an important prerequisite for further image manipulation,has attracted many scholars to study it,resulting in the proposal of countless segmentation algorithms.Among these proposed image segmentation algorithms,image threshold segmentation is a classical segmentation algorithm,but the traditional method of solving thresholds is the exhaustive method,which increases the computational and time complexity of the algorithm exponentially as the number of solved thresholds increases.Therefore to improve the speed of solving thresholds group intelligence algorithm is applied to the field.In this paper,the improved mayfly algorithm and sparrow search algorithm are applied to image threshold segmentation aiming to ensure the segmentation accuracy while speeding up the segmentation,and the main research is as follows.(1)Improved mayfly algorithm and its application in image threshold segmentation method.a)The improvement of the mayfly algorithm mainly includes two parts,one is to use the chaos algorithm in the initial stage of the mayfly algorithm to improve the quality of the initial population,and the other is to use the two-way dimension-by-dimension Tent chaotic disturbance in the algorithm search stage to enhance the search ability of the algorithm and jump out of the local area The optimal ability,combined with two aspects,proposed the chaotic mayfly algorithm CMA.Through the comparative experimental results of the algorithm in 16 benchmark functions,it can be proved that the proposed CMA algorithm can effectively improve the problem of falling into prematurity,and has higher evaluation accuracy,faster convergence speed and stronger stability than the other three algorithms.b)Combining the improved mayfly algorithm with the multi-threshold image segmentation method.The image segmentation experiment analyzes the value of each threshold value,SSIM,PSRN and variance of each algorithm under different thresholds.Through the analysis,it is proved that the CMA algorithm is an improved and successful algorithm,which can greatly improve the accuracy of the segmentation.Speed up image segmentation.(2)Improved sparrow algorithm and its application in image threshold segmentation methods.a)The improvement of the sparrow algorithm has two main parts,one is to use the reverse elite learning strategy in the initial phase of the sparrow algorithm to improve the quality of the initial population,and the other is to use the bidirectional dimension-by-dimension Tent chaos perturbation in the search phase of the algorithm to enhance the search ability of the algorithm and the ability to jump out of the local optimum.Considering that the name of this improved chaos perturbation sparrow algorithm based on the reverse elite learning strategy is too long and does not well represent the bidirectional dimension-by-dimension Tent chaos perturbation used in this paper,the improved sparrow algorithm is named the new chaos perturbation sparrow algorithm NCPSA after comprehensive consideration.12 benchmark functions are used to compare the experiments to prove that the proposed NCPSA can effectively improve the problem of falling into premature The proposed NCPSA can effectively improve the problem of falling into prematureness and has higher accuracy,faster convergence and greater stability than the other three algorithms in solving the problem.b)Combining the improved sparrow algorithm with a multi-threshold image segmentation method.The image segmentation experiments analyzed the values of each threshold taking,SSIM,PSRN and variance of each algorithm in different thresholding cases,and the analysis proved to demonstrate that the NCPSA algorithm is an improved and successful algorithm that can substantially accelerate the image segmentation speed on the basis of maintaining the segmentation accuracy.c)The CMA and NCPSA were compared and analyzed,and the experiments proved that the improved sparrow algorithm is more adapted to high-dimensional single-peak and multi-peak test functions,while the improved mayfly algorithm is relatively more adapted to low-dimensional multi-peak functions,and the image segmentation experiments prove that both algorithms have successful improvements on image threshold segmentation.24 figures,9 tables,52 references.