Code.mat

Image divided into nine small images
image=rgb2gray(imread(imgetfile));
[x,y]=size(image);
image1=image( 1     : x/3 ,   1       :   y/3);
image2=image( 1     : x/3 ,   y/3:   2*y/3);
image3=image( 1     : x/3 ,   2*y/3   :   y);
image4=image( x/3 : 2*x/3   ,   1       :   y/3);    
image5=image( x/3 : 2*x/3   ,   y/3   :   2*y/3);
image6=image( x/3 : 2*x/3   ,   2*y/3   :   y);
image7=image( 2*x/3 : x   ,   1       :   y/3);    
image8=image( 2*x/3 : x   ,   y/3   :   2*y/3);
image9=image( 2*x/3 : x   ,   2*y/3   :   y);
imwrite(image1,('image1.jpg'));
imwrite(image2,('image2.jpg'));
imwrite(image3,('image3.jpg'));
imwrite(image4,('image4.jpg'));
imwrite(image5,('image5.jpg'));
imwrite(image6,('image6.jpg'));
imwrite(image7,('image7.jpg'));
imwrite(image8,('image8.jpg'));
imwrite(image9,('image9.jpg'));

Images enhancement:

function [Image]=enhancement(image)
background = imopen(im,strel('disk',12));
image1=image-background;
Image= imadjust(image1);

Statistical features’ extraction:

function feature=feature statistical(image)
feature(1) =mean (mean (image));
feature(2) =std (sum (image));
feature(3) =std (sum (image'));

 Fourier features’ extraction:

Function feature=feature fourier(image)
k=[344 348 349 355 516 520 521 525 528 532 ];
image=abs(fftshift(fft2(image)));
[x,y]=size(image);
feature (1)=max(max(image));
image-dr1=image(:,(y/2));
image-dr2=image((x/2),:);
feature (2) =max(image-dr1(k(1):k(2)));
feature (3) =max(image-dr1(k(3):k(4)));
feature (4)=max(image-dr2(k(5):k(6)));
feature (5)=max(image-dr2(k(7):k(8)));
feature (6)=max(image-dr2(k(9):k(10)));



 Matlab code for classification using statistical features only:

 First approach 

clear
clc
t=0;
target=[];
no_folders=11;
no_images=10;
for no_folder=1:no_folders
for no_image_loop =1:no_images
image=imread(strcat(image,int2str(no_folder),'\1 (',int2str(no_image_loop),').JPG' ));
image=enhansment(image);
feature=feature_statistical(image);
t=t+1;
feature(t,:)=feature;
tt=zeros(no_folders,no_images);
tt(no_folder,:)=1;
end
target=[target tt];
end
input=feature';
net = patternnet([25 25],'trainrp');
net = train(net,input,target);
outputs = net(input);
 plotconfusion(target,outputs)

Second approach

% --- Executes on button press in classification_S.
function calssification_S_Callback(hObject, eventdata, handles)
% hObject    handle to defect_or_free_S (see GCBO)
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
close all
clear
clc
% number folders 
End folder=input('number folders = ');
number folder=end folder;
tt=0;
targets=[];
for no folder=1:end_folder
% number of images 
    No image = input(strcat('number image folder ',int2str(no folder) ,'='));
for no image loop = 1:no_image
           % read image
im=imread(strcat(image,int2str(no folder),'\1 (',int2str(no image loop),').JPG'));
% enhansment image
[im]=enhansment(im);
% feature extraction 
feature=feature statistical(im);
tt=tt+1;
inputs(:,tt)=feature;
t=zeros(number folder,no image);
t(no_folder,:)=1;
end
targets=[targets t];
end
 save('targets_inputs.mat','targets','inputs');

 Matlab code for classification using Fourier features only:

First approach

Clear
clc 
t=0;
target=[];
no folders=11;
no images=10;
for no folder=1:no folders
for no image loop =1:no_images
a=imread(strcat(image,int2str(no_folder),'\1 (',int2str(no image loop),').JPG' ));
a2=enhansment(a);
feature=feature fourier (a2);
t=t+1;
feature(t,:)=feature;
tt=zeros(no folders,no images);
tt(no folder,:)=1;
end
target=[target tt];
end
input=feature';
net = patternnet([25 25],'trainrp');
net = train(net,input,target);
outputs = net(input);
 plotconfusion(target,outputs)
  
Second approach

% --- Executes on button press in classification_F.
function classification _F_Callback(hObject, eventdata, handles)
% hObject    handle to defect_or_free_F (see GCBO)
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
close all
clear
clc
% number folders 
end_folder=input('number folders = ');
number_folder=end_folder;
tt=0;
targets=[];
for no folder=1:end folder
% number of images 
    No image = input(strcat('number image folder ',int2str(no folder) ,'='));
for no image loop = 1:no_image
           % read image
im=imread(strcat(image,int2str(no_folder),'\1(',int2str(no image loop),').JPG'));
% enhansment image
[im]=enhansment(im);
% feature extraction 
Feature=feature fourier(im);
tt=tt+1;
inputs(:,tt)=feature;
t=zeros(number folder,no image);
t(no folder,:)=1;
end
targets=[targets t];
end
 save('targets_inputs.mat','targets','inputs');

Matlab code for classification using Statistical analysis and Fourier features:

Frist approach

clear
clc
t=0;
target=[];
no folders=11;
no images=10;
for no folder=1:no_folders
for no image loop =1:no_images
a=imread(strcat(image,int2str(no folder),'\1 (',int2str(no image loop),').JPG' ));
a2=enhansment(a);
feature2=feature fourier(a2);
feature1=feature statistical(a2);
t=t+1;
feature(t,:)=[feature1 feature2];
tt=zeros(no folders,no images);
tt(no folder,:)=1;
end
target=[target tt];
end
input=feature';
 net = patternnet([25 25],'trainrp');
net = train(net,input,target);
view(net)
outputs = net(input);
 plotconfusion(target,outputs)


Second approach

% --- Executes on button press in classification_S_F.
function classification _S_F_Callback(hObject, eventdata, handles)
      % hObject    handle to defect_or_free_S_F (see GCBO)
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
% % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % 
close all
clear
clc
% number folders 
end_folder=input('number folders = ');
number_folder=end_folder;
tt=0;
targets=[];
for no_folder=1:end_folder
% number of images 
    no_image = input(strcat('number image folder ',int2str(no_folder) ,'='));
for no_image_loop = 1:no_image
           % read image
im=imread(strcat(image,int2str(no_folder),'\1 (',int2str(no_image_loop),').JPG'));
% enhansment image
[im]=enhansment(im);
% feature extraction 
Feature1=feature statistical(im);
Feature2=feature fourier(im);
tt=tt+1;
inputs(:,tt)=[ Feature1 Feature2];
t=zeros(number_folder,no_image);
t(no_folder,:)=1;
end
targets=[targets t];
end
 save('targets_inputs.mat','targets','inputs');




Matlab code for locations of Fourier spectrum peaks:

   close all
clear
clc
im=imread(imgetfile);
% enhansment image
[im]=enhansment(im);
J=abs(fftshift(fft2(im)));
[x,y]=size(J);
J((x/2),(y/2))=0; 
 A=J(343:373 , 515);
B=J(343 , 515:545);
 A1=A';
B1= B;
j=1;
    for i=1:31
if   0.35*A1(i)>mean (A1)
    AA(j,1)=A1(i);
    AA(j,2)=i;   
end
    end
    [x,y]=size(AA); 
   for i=2:x     
 if AA(i,2)==AA(i-1,2)+1
     n=AA(i,1);
     m=AA(i-1,1);
    if n>m
        AA(i-1,:)=0;
    else
       AA(i,:)=0;
    end 
 end
   end 
   j=1;
    for i=1:31
if   0.35*B1(i)>mean(B1)
    BB(j,1)=B1(i);
    BB(j,2)=i;   
end
    end
    [x,y]=size(BB); 
   for i=2:x     
 if BB(i,2)==BB(i-1,2)+1
     n=BB(i,1);
     m=BB(i-1,1);
    if n>m
        BB(i-1,:)=0;
    else
       BB(i,:)=0;
    end 
 end
   end    
     figure; plot(A1)
figure;plot(B1)
    AA(:,2)
    BB(:,2)

Create  neural network 

function [net]=Neural_Network_mohy(inputs,targets)
% Solve a Pattern Recognition Problem with a Neural Network
% Script generated by NPRTOOL
% Created Fri Jun 21 15:38:32 EEST 2013
% This script assumes these variables are defined:
%   inputs - input data.
%   targets - target data.
 % Create a Pattern Recognition Network
load('targets_inputs.mat')
hiddenLayerSize = 25;
net = patternnet([hiddenLayerSize hiddenLayerSize]);
 % Choose Input and Output Pre/Post-Processing Functions
% For a list of all processing functions type: help nnprocess
net.inputs{1}.processFcns = {'removeconstantrows','mapminmax'};
net.outputs{2}.processFcns = {'removeconstantrows','mapminmax'};
  % Setup Division of Data for Training, Validation, Testing
% For a list of all data division functions type: help nndivide
net.divideFcn = 'dividerand';  % Divide data randomly
net.divideMode = 'sample';  % Divide up every sample
net.divideParam.trainRatio = 80/100;
net.divideParam.valRatio = 10/100;
net.divideParam.testRatio = 10/100;
%  trainrp RPROP backpropagation.
%    trainrp is a network training function that updates weight and bias
%    values according to the resilient backpropagation algorithm (RPROP).
% For help on training function 'trainrp' type: help trainrp
% For a list of all training functions type: help nntrain
net.trainFcn = 'trainrp'; 
% Choose a Performance Function
% For a list of all performance functions type: help nnperformance
net.performFcn = 'mse';  % Mean squared error
 % Choose Plot Functions
% For a list of all plot functions type: help nnplot
net.plotFcns = {'plotperform','plottrainstate','ploterrhist', ...
  'plotregression', 'plotfit'};
 % Train the Network
[net,tr] = train(net,inputs,targets);
 % Test the Network
outputs = net(inputs);
errors = gsubtract(targets,outputs);
performance = perform(net,targets,outputs);
 % Recalculate Training, Validation and Test Performance
trainTargets = targets .* tr.trainMask{1};
valTargets = targets  .* tr.valMask{1};
testTargets = targets  .* tr.testMask{1};
trainPerformance = perform(net,trainTargets,outputs);
valPerformance = perform(net,valTargets,outputs);
testPerformance = perform(net,testTargets,outputs);
 % View the Network
view(net)
 % Plots
% Uncomment these lines to enable various plots.
figure, plotperform(tr)
 saveas(gcf,'1.jpg')
figure, plottrainstate(tr)
 saveas(gcf,'2.jpg')
figure, plotconfusion(targets,outputs)
 saveas(gcf,'3.jpg')
figure, ploterrhist(errors)
 saveas(gcf,'4.jpg')

Test image using Statistical analysis and Fourier features:

% --- Executes on button press in test_S_F.
function test_S_F_Callback(hObject, eventdata, handles)
% hObject    handle to test_S_F (see GCBO)
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
close all
clear
clc
load('net.mat')
im=imread(imgetfile);
imshow(im)
[im]=enhansment(im);
Feature1= feature statistical(im);
Feature2=feature fourier(im);
test=[feature1 feature2];
s=sim(net,test’);
[p,q]=max(s);
q

Test image using Statistical analysis:

% --- Executes on button press in test_S.
function test_S_Callback(hObject, eventdata, handles)
% hObject    handle to test_S (see GCBO)
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
close all
clear
clc
load('net.mat')
im=imread(imgetfile);
imshow(im)
[im]=enhansment(im);
Feature=feature statistical(im);
test=feature;
s=sim(net,test’);
[p,q]=max(s);
q

Test image using Fourier features:

% --- Executes on button press in test_F.
function test_F_Callback(hObject, eventdata, handles)
% hObject    handle to test_F (see GCBO)
% eventdata  reserved - to be defined in a future version of MATLAB
% handles    structure with handles and user data (see GUIDATA)
close all
clear
clc
load('net.mat')
im=imread(imgetfile);
imshow(im)
[im]=enhansment(im);
Feature=feature fourier(im);
test= Feature;
s=sim(net,test’);
[p,q]=max(s);
q


Test any image 

Frist approach 

load('net.mat');
im=imread(imgetfile);
imshow(im)
[im]=enhansment(im);
feature1=feature statistical(im);
feature2=feature fourier(im);
test=[feature1 feature2];
s=sim(net,test’);
[p,q]=max(s);
if q==1 
    display ('kont')
elseif q==2
    display('float warp')
    elseif q==3
    display('light beat')
    elseif q==4
    display('heavy beat')
    elseif q==5
    display('missing picks')
    elseif q==6
    display('double end')
    elseif q==7
    display('hole')
    elseif q==8
    display('stain')
    elseif q==9
    display('double pick')
    elseif q==10
    display('free')
else
    display('big kont')
end

Second approach

clear
clc
a1=load('net 11.mat');
a2=load('net 22.mat');
a3=load('net 33.mat');
a4=load('net 41.mat');
a5=load('net 51.mat');
im=imread(imgetfile);
imshow(im)
[im]=enhansment(im);
Feature=feature statistical(im);
test=feature;
s=sim(a1.net,test’);
[p,q]=max(s);
if q==2
  display('free')
else
    display('defect')
feature = feature fourier(im);
test= feature;
s=sim(a2.net,test’);
[p,q]=max(s);
if q==1
 display('area')
feature1=feature statistical(im);
feature2=feature fourier(im);
test=[feature1 feature2];
s=sim(a3.net,test’);
[p,q]=max(s);
 if q==1
     display('kont')
 elseif q==2
     display('hole')
 else
     display('stain')
 end    
elseif q==2
    display('warp')
feature=feature statistical(im);
test=feature;
s=sim(a4.net,test’);
[p,q]=max(s);
if q==1 
    display('Double end')
else
    display('Float warp')
end
else
    display('weft')
feature=feature statistical(im);
test=feature;
s=sim(a5.net,test’);
[p,q]=max(s);
if q==1
    display('ligth beat')
elseif q==2
    display('heavy beat')
elseif q==3
    display('missing picks')
else
    display('Double pick')
end
end
end