Research Of Adaptive Chaos Particle Swarm Optimization Algorithm And Application In Image Segmentation

Particle Swarm Optimization (PSO) algorithm is a heuristic swarm intelligence optimization algorithm which derives from birds foraging. Because of its fast convergence, simple thoughts and easy implementation, PSO algorithm has gotten extensive attention and in-depth research. PSO algorithm has been applied in function optimization, neural network training, fuzzy control and other industrial technology fields successfully.The domestic and foreign research scholars have done a lot of researches in allusion to the algorithm’s defects of false convergence to the local extreme value, slow rate of convergence in the later iteration and premature problem and have proposed improved method. Improved methods can be divided into three categories:the improvement of algorithm parameters, the improvement of algorithm evolution process and the algorithm’s combining with other algorithms. Improved algorithm speeds up the convergence and inhibits premature convergence of the algorithm as much as possible..In this thesis, according to the shortage of the algorithm false convergence to some local extreme value, slow rate of convergence in the later iteration and low accuracy of optimization results, the thesis proposes the following improvements:(1) An improved adaptive Particle Swarm Optimization (IAPSO) algorithm, which dynamically modulats the inertia weight in the standard PSO algorithm according to the distance between a particle and the globally best particle’s position found so far, and combines with a position updated equation with a mobility factor.(2) On the base of reserving the advantages of the IAPSO algorithm, also to overcome the disadvantages of the premature convergence during the later computation period of PSO algorithm, the thesis proposes an adaptive chaos particle swarm optimization (ACPSO) algorithm with a chaos optimization method. It initializes the position and velocity of the particles using the chaso variables. The particles are mutated by Iterative Chaotic Map with Infinite Collapses (ICMIC), and the best part of the particles chosen from the current population realizes chaos optimization. Finally, ACPSO algorithm is applied to the image segmentation.The thesis does a comparison test experiment about improved algorithm using standard test functions. The rate of convergence of ACPSO algorithm is1.825-20.06times than that of IAPSO algorithm, is4.63-32.55times than that of basic PSO algorithm; The rate of convergence of IAPSO algorithm is3.27-14.66times than that of IAPSO algorithm, is1.51-2.05times than that of PSO algorithm with constriction factor.The optimization accuracy of ACPSO algorithm is at least75%success rate to converge to the optimal solution.For image segmentation based on ACPSO algorithm and two-dimensional maximum fuzzy Shannon entropy threshold segmentation, average segmentation standard deviation decreases by2.27%than the original algorithm, and the speed of convergence increases by6.94times.