模式识别matlab程序

模式识别的一些matlab程序

最小错误率Bayes分类器的设计与检验 clc

clear

X1 = 10000;

MU1 = 2.0;

SIGMA1 = 0.2;

Y1 = normrnd(MU1, SIGMA1, X1, 1);

X2 = 5000;

MU2 = 1.0;

SIGMA2 = 0.2;

Y2 = normrnd(MU2, SIGMA2, X2, 1);

Y = [Y1;Y2];

Pw1 = X1 / (X1+X2);

Pw2 = X2 / (X1+X2);

T1 = find(normpdf(Y1,MU1,SIGMA1)*Pw1 > normpdf(Y1,MU2,SIGMA2)*Pw2); T2 = find(normpdf(Y2,MU2,SIGMA2)*Pw2 > normpdf(Y2,MU1,SIGMA1)*Pw1); et = (X1-length(T1)+X2-length(T2)) / (X1+X2);

t = fsolve('fun1', 1);

e = quadl('fun2', -10000, t) + quadl('fun3', t, 10000);

%fun1

function e = fun1(x)

MU1 = 2.0;

SIGMA1 = 0.2;

MU2 = 1.0;

SIGMA2 = 0.2;

e = normpdf(x,MU1,SIGMA1)*2/3 - normpdf(x,MU2,SIGMA2)*1/3; %fun2

function y = fun2(x)

MU1 = 2.0;

SIGMA1 = 0.2;

MU2 = 1.0;

SIGMA2 = 0.2;

y = normpdf(x,MU1,SIGMA1)*2/3;

%fun3

function y = fun3(x)

MU1 = 2.0;

SIGMA1 = 0.2;

MU2 = 1.0;

SIGMA2 = 0.2;

y = normpdf(x,MU2,SIGMA2)*1/3;

窗函数法高斯分布

clc

clear

load('TestData.mat')

Y = [Y1;Y2];

%hist([Y1;Y2],x)

hn = 0.01;

x = 0:0.01:3;

%y = 1/15000*sum(1/hn*normpdf((x-Y),MU,SIGMA));

for i = 1:3/0.01+1

y(i) = 1/15000*sum(1/hn*normpdf((x(i)-Y)/hn, 0, 1)); end

plot(x,y)

hold on

MU1 = 2.0;

SIGMA1 = 0.2;

MU2 = 1.0;

SIGMA2 = 0.2;

z = normpdf(x,MU1,SIGMA1)*2/3 + normpdf(x,MU2,SIGMA2)*1/3; plot(x,z,'r')

近邻法高斯分布

clc

clear

load('TestData.mat')

Y = [Y1;Y2];

Y = sort(Y);

x = 0:0.01:3;

kn = 100;

for i = 1:3/0.01+1

j = 1;

while 1

if abs(Y(j)+Y(j+kn-1)-2*x(i)) < abs(Y(j+1)+Y(j+kn)-2*x(i)) break;

else

j = j+1;

end

if j == 15000-kn+1

break;

end

end

y(i) = kn/15000/(Y(j+kn-1)-Y(j));

end

plot(x, y)

hold on

MU1 = 2.0;

SIGMA1 = 0.2;

MU2 = 1.0;

SIGMA2 = 0.2;

z = normpdf(x,MU1,SIGMA1)*2/3 + normpdf(x,MU2,SIGMA2)*1/3; plot(x,z,'r')

Fisher线性变换

clc

clear

X1 = 5;

One1 = ones(X1, 1);

SIGMA1 = 0.2;

Y1 = normrnd([One1.*2 One1.*3], SIGMA1, X1, 2);

X2 = 5;

One2 = ones(X2, 1);

MU2 = 3.0;

SIGMA2 = 0.2;

Y2 = normrnd([One2.*3, One2.*2], SIGMA2, X2, 2);

plot(Y1(:,1), Y1(:,2), 'r*', Y2(:,1), Y2(:,2), 'bo')

m1 = mean(Y1);

m2 = mean(Y2);

S1 = (Y1 - One1 * m1)'*(Y1 - One1 * m1);

S2 = (Y2 - One2 * m2)'*(Y2 - One2 * m2);

sw = S1+S2;

w = inv(sw)*(m1 - m2)';

Y = [Y1;Y2];

z = Y*w;

hold on

t = z*w'/norm(w)^2;

plot(t(:,1), t(:,2))

for i = 1:X1+X2

plot([Y(i,1) t(i,1)], [Y(i,2) t(i,2)], '-.')

end

axis([-1 3.5 -1 3.5])

grid

一种基于最近邻优先的知识聚类算法 clear all

clc

I = imread('InPut.bmp');

M = rgb2gray(I);

clear I

I = im2bw(M,254/255);

clear M

Igl = I;

%进行第一次聚类

Igl = process(Igl, 7);

subplot(1,2,1)

subimage(I)

subplot(1,2,2)

subimage(Igl)

%第一次

…

%第...次

% process

function Igl = process(I, r)

[A B] = find(I == 0);

A = round(mean(A));

B = round(mean(B));

[X, Y] = FindNextPoint(I, A, B, 0)

[Gx, Gy, Igl] = FindGroupPoint(I, X, Y, r, 0);

% FindNextPoint

function [X, Y] = FindNextPoint(I, A, B, x);

[a, b] = size(I);

temp = max([A, B, a - A, b - B]);

for i = 1 : temp

Q = I(A-i:A+i,B-i:B+i);

[X, Y] = find(Q == x);

if length(X) ~= 0

X = X(1) + A-i - 1;

Y = Y(1) + B-i - 1;

break;

end

end

% FindGroupPoint

function [Gx, Gy, Igl] = FindGroupPoint(I, X, Y, r, x)

Igl = I;

while(1)

temp = length(X);

if(temp == 0)

Gx = 0;

Gy = 0;

break;

end

T = find(Igl == x);

pp = length(T)

[X, Y, Igl] = FindNearPoint(Igl, X, Y, r, x); end