An Improved Differential Evolution Algorithm Based On Population Diversity And Its Application In Image Registration

The optimization problem has always been closely related to the production and life of man kind.since the dawn of mankind.On the road of long-term exploration related to the solution to the problem,humans are optimizing their own logical methods constantly.In the natural space of existence,people are learning to explore the experience of survival,as well.There are countless laws and rule around nature.People became aware and master the rule gradually in the civilization thousands of years.With the development of modern society,due to birth of the electronic computer,much optimization problem can be defined as the computer simulated computation.That is bionic intelligence simulation.It can be used as optimization problem solving and the practical application.Differential evolutionary algorithm,with the help of Darwin's "Survival of the fittest",is the important branch of the bionic intelligence simulation.Study shows that differential evolutionary algorithm outshine other random algorithm about the rate of convergence and stability.Due to the simple operation of the differential evolution algorithm and the few parameters,many researchers put forward a large number of research on optimized differential evolutionary algorithms.Three basic operations and mathematical models of the variation,crossover and selection of the differential evolution algorithm show,the search efficiency is obviously determined by its control parameter setting.In recent years,researchers have proposed various evolutionary strategies about the adaptive and constrained concept of the control parameters around the algorithm.However,the generalization ability of the differential evolution algorithm is not good at present,and it usually compare the overall effect of the algorithm refer to some or several specific problems.Population diversity is an important concept mentioned in particle swarm optimization.This concept is also applied to other group intelligence calculations.In many researches,the parameter adaptability of the algorithm and the addition of constraints have always been an important strategy of optimization algorithm.This dissertation starts with the optimization problem,introduces the development of the optimization problem and the development status abroad,which introduces differential evolution algorithm and population diversity that need to be studied as following.This dissertation introduces the definition of population diversity and its mathematical model and three basic operations of differential diversity.The three basic operations are used as the overall algorithm framework of this dissertation.Population diversity is one of five basic principles refer to particle swarm optimization's population behavior.During the optimizing process,as the search fixed area of a population.The innovation points of my dissertation are as follows:(1)For the JADE algorithm involved in this article,with the In-depth analysis of the single variation strategy selection problem of the middle parameters and updated randomness,My dissertation try to put forward a solution.To solve the generalization problem of JADE algorithm,SHADE algorithm use list of parameters index,to determine the parameters of the next individual according to the records of the successful individual.and the updating of the certain position parameters in the index list.This work is the basis of the overall architecture of this algorithm.(2)In view of the existing difference algorithm,once the population size is fixed,the size of the search range is fixed,My dissertation In this paper,on the parameter p that controls the size of the search scope,according to the global search of population diversity and the experimental experience in each generation search.After normalization of population diversity,as a constraint factor in the search scope,the optimal solution can be found in the dynamic search range after ensuring that each individual can be constrained by the conditions of diversity,which can obtain the global optimal solution,the local optimal approximation is guaranteed.(3)The dissertation's application experiment of image registration is carried out to verify the practicability of the algorithm.In addition,compared with many existing advanced algorithms,the overall accuracy of the algorithm is superior to that of other classical algorithms.