PSO的一维最大熵法阈值分割

Halcom · 发表于 2019-10-10 21:53:52

PSO的一维最大熵法阈值分割
链接：https://pan.baidu.com/s/1ChO4xGObCiTMqRT_s_qDpg&shfl=sharepset 提取码：zoad
具体链接在halcom.cn论坛，联系人QQ：3283892722
该论坛是一个学习交流平台，我会逐一的和大家分享学习。
欢迎大家录制视频，你可在论坛进行打赏分享。
视频专用播放器：http://halcom.cn/forum.php?mod=viewthread&tid=258&extra=page%3D1\

clc,clear,close all
warning off
format longG
tic;
% 2个阈值，实现3分类分割
I = imread('car.bmp');
% I = imread('screen.jpg');
I = imresize(I, [256,256]);
if(size(I,3)>1)
im = rgb2gray(I);
else
im = I;
end
figure('color',[1,1,1]), imshow(im,[])
%% PSO 参数
c1 = 1.4995; % 学习因子
c2 = 1.4995; % 学习因子
Vmin = -5; % 最小速度
Vmax = 5; % 最大速度
maxiter = 30; % 迭代次数
sizepop = 20; % 种群数量
popmin = 1; popmax = 255; % x1
Lmax = 256;
% 2个阈值，实现3分割
level = 3;
nvar = level-1;
%% 初始化种群
for i=1:sizepop
x1 = fix( popmin + (popmax-popmin)*rand );
x2 = fix( popmin + (popmax-popmin)*rand );
pop(i,1) = x1;
pop(i,2) = x2;
pop(i,:)=sort(pop(i,:));
fitness(i) = H1_segFun2( im, pop(i,:) );
V(i,1) = 0;
V(i,2) = 0;
end
% 记录一组最优值
[bestfitness,bestindex]=max(fitness);
zbest=pop(bestindex,:); % 全局最佳
gbest=pop; % 个体最佳
fitnessgbest=fitness; % 个体最佳适应度值
fitnesszbest=bestfitness; % 全局最佳适应度值
wmax = 0.9; wmin = 0.4;
% 迭代寻优
for i=1:maxiter
for j=1:sizepop
% 自适应权重1
% w = wmin + (wmax-wmin)*(fitnessgbest(j)-min(fitness))/( max(fitness)-min(fitness) );
% 自适应权重2
% w = wmin - (wmax-wmin)*(fitnessgbest(j)-min(fitness))/( mean(fitness)-min(fitness) );
% 自适应权重3
if fitnessgbest(j)<=mean(fitness)
w = wmin - (wmax-wmin)*(fitnessgbest(j)-min(fitness))/( mean(fitness)-min(fitness) );
else
w = wmax;
end
% 速度更新
V(j,:) = w*V(j,:) + c1*rand*(gbest(j,:) - pop(j,:)) + c2*rand*(zbest - pop(j,:));
% V--范围约束
V(j, find(V(j,:)>Vmax)) = Vmax;
V(j, find(V(j,:)<Vmin)) = Vmin;
% 个体更新
pop(j,:) = fix( pop(j,:) + 0.5 * V(j,:) );
% pop 越界限制
pop(j, find(pop(j,:)>popmax)) = popmax;
pop(j, find(pop(j,:)<popmin)) = popmin;
pop(j,:) = sort(pop(j, :));
% 适应度更新
fitness(j) = H1_segFun2(im, pop(j,:));
% 比较个体间比较
if fitness(j)>fitnessgbest(j)
fitnessgbest(j) = fitness(j);
gbest(j,:) = pop(j,:);
end
if fitness(j)>bestfitness
bestfitness = fitness(j);
zbest = pop(j,:);
end
end
fitness_iter(i) = bestfitness;
end
toc ;
times = toc;
fprintf('\n')
disp(['计算时间 Time = ', num2str(times) ])
fprintf('\n')
zbest = sort(zbest);
disp(['最优解 ', num2str(zbest)])
fprintf('\n')
figure('color',[1,1,1])
plot(fitness_iter,'ro-','linewidth',2)
% 显示
Iout = imageGray( im, zbest );
figure('color',[1,1,1])
imshow(Iout,[])
colormap jet;
axis equal

复制代码

适应度函数：

function maxH = H1_segFun2(I, t)
% 最原始的一维最大熵法，进行图像的3分割,level=3,将图像分为3部分
[n_count, x_value] = imhist(I(:,:,1)); % 第一通道来分割
Nt = size(I,1)*size(I,2);
Lmax = 256; % 最大灰阶, uin8
for i=1:Lmax
p(i) = n_count(i)/Nt; % 概率分布
end
% H0 = -1* sum( p(1:t).*log(p(1:t) + eps) );
% HB = -1* sum( p(t+1:Lmax).*log(p(t+1:Lmax) + eps) );
% maxH = H0+HB; % 一维最大熵
% 改进
% p1 = log( sum(p(1:t)) + eps );
% H0 = sum( p(1:t).*log(p(1:t) + eps) )./ ( sum(p(1:t)) + eps );
% p2 = log( sum(p(t+1:Lmax)) + eps );
% HB = sum( p(t+1:Lmax).*log(p(t+1:Lmax) + eps) )./ ( sum(p(t+1:Lmax)) + eps );
% maxH = p1+p2-H0-HB; % 一维最大熵
% % 3分割
% t1 = t(1);
% t2 = t(2);
% p1 = log( sum(p(1:t1)) + eps );
% p2 = log( sum(p(t1+1:t2)) + eps );
% p3 = log( sum(p(t2+1:Lmax)) + eps );
% p4 = sum( p(1:t1).*log(p(1:t1) + eps) )./ ( sum(p(1:t1)) + eps );
% p5 = sum( p(t1+1:t2).*log(p(t1+1:t2) + eps) )./ ( sum(p(t1+1:t2)) + eps );
% p6 = sum( p(t2+1:Lmax).*log(p(t2+1:Lmax) + eps) )./ ( sum(p(t2+1:Lmax)) + eps );
% maxH = p1+p2+p3-p4-p5-p6; % 一维最大熵
% 3分割
% t1 = t(1);
% t2 = t(2);
p1 = log( sum(p(1:t(1))) + eps );
p2 = log( sum(p(t(1)+1:t(2))) + eps );
p3 = log( sum(p(t(2)+1:Lmax)) + eps );
p4 = sum( p(1:t(1)).*log(p(1:t(1)) + eps) )./ ( sum(p(1:t(1))) + eps );
p5 = sum( p(t(1)+1:t(2)).*log(p(t(1)+1:t(2)) + eps) )./ ( sum(p(t(1)+1:t(2))) + eps );
p6 = sum( p(t(2)+1:Lmax).*log(p(t(2)+1:Lmax) + eps) )./ ( sum(p(t(2)+1:Lmax)) + eps );
maxH = p1+p2+p3-p4-p5-p6; % 一维最大熵

复制代码

参考：【32】一维最大熵法进行2分割和3分割（MATLAB）

		自动登录	找回密码
密码			立即注册