www.gusucode.com > 《图像配准技术及其Matlab编程实现》--源码程序 > 《图像配准技术及其Matlab编程实现》/match/GA.m

    function [x,y,max_value]=GA

global M T Gen NextGen

pr=initialization();  %初始化遗传算法参数
[Gen NextGen]=initGeneration(pr);  %产生父代及子代
for i=1:T
    [Fitness x1 x2]=CalculateFitness(pr);
    [OrderFitness,Index]=sort(Fitness);  %将适应度从小到大进行排序
    MaxFitness=OrderFitness(M);  %找出适应度的最大值
    AvgFitness=mean(Fitness);  %求出适应度的平均值
    Diff=MaxFitness - AvgFitness;
    Best_x1=x1(Index(M));
    Best_x2=x2(Index(M));
    BetterFit=SaveElitist(Index,pr);  %保护精英个体
    Select(Fitness,pr);  %选择操作
    Cross(OrderFitness,AvgFitness,Diff,pr);  %交叉操作
    Mutation(OrderFitness,MaxFitness,AvgFitness,Diff,pr);  %变异操作
    RestoreElitist(BetterFit,pr);  %恢复精英个体
    Gen=NextGen;  %产生新一代
end

max_value=MaxFitness;  %输出结果
x=Best_x1;
y=Best_x2;

function pr=initialization()
global M T
M=80;   %种群大小
T=200;  %进化代数
pr.ChromLong=22;  %染色体长度（二进制编码）
pr.Pcross1=0.9;   %交叉概率常数
pr.Pcross2=0.6;
pr.Pmutation1=0.1;  %变异概率常数
pr.Pmutation2=0.001;
pr.ElitistNumber=5;  %精英个体数量
pr.VaribleNumber=2;  %变量个数
pr.Lower1=-10;       %设置变量定义域
pr.Upper1=10;
pr.Lower2=-10;
pr.Upper2=10;

%初始化父代及子代
function [Gen NextGen]=initGeneration(pr)
global M Gen NextGen
CL=pr.ChromLong;
VN=pr.VaribleNumber;
Gen=round(rand(M,CL*VN));  %随机产生父代
NextGen=zeros(M,CL*VN);    %初始化子代

%计算适应度
function [Fitness x1 x2]=CalculateFitness(pr)
global M Gen
CL=pr.ChromLong;
Upper1=pr.Upper1;
Upper2=pr.Upper2;
Lower1=pr.Lower1;
Lower2=pr.Lower2;
for s=1:M
    Individual=Gen(s,:);
    decimal_x1=0;
    decimal_x2=0;
    Chrom_1=Individual(1:CL);  %对x1进行编码
    for i=1:CL
        decimal_x1=decimal_x1+Chrom_1(i)*2^(i-1);
    end
    x1(s)=(Upper1-Lower1)*decimal_x1/(2^CL-1)+Lower1;
    
    Chrom_2=Individual(CL+1:1:2*CL);  %对x2进行解码
    for i=1:CL
        decimal_x2=decimal_x2+Chrom_2(i)*2^(i-1);
    end
    x2(s)=(Upper2-Lower2)*decimal_x2/(2^CL-1)+Lower2;
    Fitness(s)=objFunction(x1(s),x2(s));  %计算适应度
end

%选择操作，采用轮盘赌选择法
function Select(Fitness,pr)
global M Gen NextGen
BFN=pr.ElitistNumber;
SumFitness=sum(Fitness);  %计算适应度和
for i=1:(M-BFN)
    Rd=rand;
    s=1;
    temp=0;
    while(temp<Rd) & (s<=M)
        temp=temp+Fitness(s)/SumFitness;  %计算积累概率
        s=s+1;
    end
    if s==1
        s=1;
    else
        s=s-1;
    end
    NextGen(i,:)=Gen(s,:);
end

%交叉操作
function Cross(OrderFitness,AvgFitness,Diff,pr)
global M NextGen
CL=pr.ChromLong;
VN=pr.VaribleNumber;
Pc1=pr.Pcross1;
Pc2=pr.Pcross2;
EN=pr.ElitistNumber;
for i=1:2:(M-EN)
    Rn=rand;
    if OrderFitness(i+1)>=AvgFitness  %计算自适应交叉概率
        pcross=Pc1-(Pc1-Pc2)*(OrderFitness(i+1)-AvgFitness)/Diff;
    else
        pcross=Pc1;
    end
    if pcross>Rn  %满足交叉条件时，进行交叉运算
        CrossPoint=ceil(VN*CL*Rn);
        if or(CrossPoint==0,CrossPoint>=VN*CL)
            continue;
        end
        for j=CrossPoint:VN*CL  %随机交换部分染色体基因
            temp=NextGen(i,j);
            NextGen(i,j)=NextGen(i+1,j);
            NextGen(i+1,j)=temp;
        end
    end
end

%变异操作
function Mutation(OrderFitness,MaxFitness,AvgFitness,Diff,pr)
global M NextGen
EN=pr.ElitistNumber;
CL=pr.ChromLong;
VN=pr.VaribleNumber;
Pm1=pr.Pmutation1;
Pm2=pr.Pmutation2;
for i=1:(M-EN)
    for j=1:VN*CL
        if OrderFitness(i)>=AvgFitness  %计算自适应变异概率
            pmutation=Pm1-(Pm1-Pm2)*(MaxFitness-OrderFitness(i))/Diff;
        else
            pmutation=Pm1;
        end
        if pmutation>rand  %满足变异条件时，进行变异运算
            if NextGen(i,j)==0
                NextGen(i,j)=1;
            else
                NextGen(i,j)=0;
            end
        end
    end
end

%保护精英个体
function BetterFit=SaveElitist(Index,pr)
global M Gen
EN=pr.ElitistNumber;
for i=1:EN
    BetterFit(i,:)=Gen(Index(M-i+1),:);
end

%恢复精英个体
function RestoreElitist(BetterFit,pr)
global M NextGen
EN=pr.ElitistNumber;
for i=1:EN
    NextGen(M-i+1,:)=BetterFit(i,:);
end

%目标函数
function F=objFunction(x1,x2)
F=0.5-((sin(sqrt(x1.^2+x2.^2))).^2-0.5)./(1+0.001*(x1.^2+x2.^2).^2);