评价网 > 科教论文 > 工学毕业论文 > 通信工程论文 > 正文

基于特征融合算法的人脸识别(一)-通信工程毕业(5)

2013-05-18 01:51
function Ps = gabor(I,w,nu,mu,Kmax,f,sig)
%I表示传入的要进行滤波的图像数据40*30
m = w(1);
n = w(2);
K = Kmax/f^nu * exp(i*mu*pi/8);
Kreal = real(K);
Kimag = imag(K);
NK = Kreal^2+Kimag^2;
Psi = zeros(m,n);
for x = 1:m
    for y = 1:n
        Z = [x-m/2;y-n/2];
        Psi(x,y) = (sig^(-2))*exp((-.5)*NK*(Z(1)^2+Z(2)^2)/(sig^2))*...
                   (exp(i*[Kreal Kimag]*Z)-exp(-(sig^2)/2));
    end
end
%图像数据与gabor的实部和xubu分别进行卷积
Imgabout = conv2(I,double(imag(Psi)),'same');
Regabout = conv2(I,double(real(Psi)),'same');

Ps = sqrt(Imgabout.*Imgabout + Regabout.*Regabout);

附录 B  对每幅人脸图像进行处理的源代码
function UltraInfo = handlesinglepic(filename)
 
if ~ischar(filename)
     UltraInfo =0;
end
imma =imread(filename);
[x y] =size(imma);  %获取图像的行列数
immqy =zeros(40,30);
G = cell(5,8);
imd =zeros(10,96); %保存gabor的取样点
for s = 1:5
   for j = 1:8
         G{s,j}=zeros(40,30);
     end
end

%%%%%%%这是奇异值分解%%%%%%%%%%%%%%%%%%%%%%%
 [U S V] =svd(double(imma));
 for ii =1:10
  A(ii,1) =S(ii,ii);
 end
%%%%%%%%%以下为提取的Gabor幅值特征%%%%%%%%%%%%%
for i =1 :40
    for j =1 :30
        xz = fix(i*x/41);
        yz = fix(j*y/31);
        immqy(i,j) =imma(xz,yz); （科教范文网 lw.AsEac.com编辑整理）
    end
end
for s = 1:5
    for j = 1:8
        G{s,j}= gabor(immqy,[40 30],(s-1),(j-1),4*pi/5,sqrt(2),3*pi/2);
    end
end
%%%%%%%%%%%%%融合前统一量纲%%%%%%%%%%%%%%%%%%%
numave =mean(imd);
numaves =mean(A);
stdnum1 =std(imd);
stdnum2 =std(A);
for ii =1:size(imd,1)
    for jj =1:size(imd,2)
        imd(ii,jj) =(imd(ii,jj) -numave(jj))/stdnum1(jj);
    end
end
for ii =1:size(A,1)
    A(ii,1) =(A(ii,1) -numaves)/stdnum2;
End
%%%%%%%%以下串行融合Gabor特征向量与该图像的奇异值%%%%%%%%
UltraInfo =[imd A];
%%%%%%%%%%%%清除多余变量%%%%%%%%%%%%%%%%%%%
clearvarible;
 
  

附录 C  最近邻分类的源代码
function mdis =mdist(a,b)
mdis =0;
for ii =1:size(a,1)
    for jj =1:size(a,2)
        mdis =mdis +(a(ii,jj)-b(ii,jj))^2;
    end
end
mdis =sqrt(mdis);

附录 D  程序执行主线源代码
clear;
close all;
clc;
%%%%%%%%%%%读取训练图像和测试图%%%%%%%%%%%%%%%%%
imagedata =cell(1,15);  %存储15幅训练图像的特征'开始基准图像文件的读取...'
for i =1:15
 imagedata{1,i} =zeros(10,97);
 temp =int2str(i);
 filename =[temp '.pgm'];
 imagedata{1,i} =handlesinglepic(filename);
 clc;
 ['第' int2str(i) '个文件处理完成... 共15幅基准图像']
end
imcompare =cell(1,75);
clc;
'开始比较文件的读取...'
filename =cell(1,75); %存储75幅测试图像
for i =1:75
 imcompare{1,i} =zeros(10,97);
 temp =int2str(100+i);

 filename{i} =[temp '.pgm'];
 imcompare{1,i} =handlesinglepic(filename{i});
 clc;
 ['第' int2str(i) '个文件处理完成... 共75幅比较图像']
end

%%%%%比较两种图像的欧式几何距离(最近邻分类)，产生结果%%%%%%%%
'正在分析分类图像...'
standardnum = 1e100;
teamnum =1;
fresult =zeros(1,75);
for ii =1:75
 standardnum = 1e100;
 for jj =1:15
  dis =mdist(imagedata{jj},imcompare{ii});
  if dis <standardnum
   standardnum =dis;
   fresult(ii) =jj;
  end
 end
end
clc;
'比较结束,共1020次比较'
%%%%%%%%%%%图形输出比较后的图像，便于比较识别率%%%%%%%
clc;
'输出原始基准图像...'
figure(1);
for ii =1:15
 subplot(3,5,ii);
 imshow(imread([int2str(ii) '.pgm']));
 title([int2str(ii) '号人物']);
end

clc;
'输出比较后的图像...'
figure(2);
text(50,50,'第1~5人的识别结果(每行为一人)');
printnum =zeros(1,15);
for jj =1:75
 if fresult(jj) >=1 & fresult(jj) <=5
  printnum(fresult(jj)) =printnum(fresult(jj)) +1;
  subplot(5,5,5*(fresult(jj)-1)+printnum(fresult(jj)));
  imshow(imread(filename{jj}));
 end
end
for i=3:5:23
 subplot(5,5,i);
 title(['以下为第 ' int2str((i-3)/5+1) ' 个人物的人脸识别结果']);
end

figure(3);
printnum =zeros(1,15);
for jj =1:75
 if fresult(jj) >=6 & fresult(jj) <=10
  printnum(fresult(jj)) =printnum(fresult(jj)) +1;
  subplot(5,5,5*(fresult(jj)-6)+printnum(fresult(jj)));
  imshow(imread(filename{jj}));
 end
end （科教范文网 lw.AsEac.com编辑整理）
for i=3:5:23
 subplot(5,5,i);
 title(['以下为第 ' int2str((i-3)/5+6) ' 个人物的人脸识别结果']);
end

figure(4);
printnum =zeros(1,15);
for jj =1:75
 if fresult(jj) >=11 & fresult(jj) <=15
  printnum(fresult(jj)) =printnum(fresult(jj)) +1;
  subplot(5,5,5*(fresult(jj)-11)+printnum(fresult(jj)));
  imshow(imread(filename{jj}));
 end
end
for i=3:5:23
 subplot(5,5,i);
 title(['以下为第 ' int2str((i-3)/5+11) ' 个人物的人脸识别结果']);
end

'完成！'