I'm developing a project faical recongnization based on Eye Distance without image database

 I'm using two parameter depth of the face using two webcams and eye distance by finding out the centroid of the eyes and class, all these three stored in .mat file and compare with test data. i'm using ecludidean distance i have developed a code which work's but accuracy is very low. Can anyone please help me. i'm trying for 6 months

here is the code -----------Depth.M--------

C = webcamlist;
if length(C)>2
    fprintf('All three WebCam is Working' );
else
    fprintf('Not All three WebCam is Working' );
end
load('webcamsSceneReconstruction.mat');
vid1 = videoinput('macvideo', 1);
vid2 = videoinput('macvideo', 2);
start([vid1 vid2]);
img1 = getsnapshot(vid1);
img2 = getsnapshot(vid2);
imwrite(img1,'sceneReconstructionLeft.jpg');
imwrite(img2,'sceneReconstructionRight.jpg');
I1 = imread('sceneReconstructionLeft.jpg');
I2 = imread('sceneReconstructionRight.jpg');
I1 = undistortImage(I1,stereoParams.CameraParameters1);
I2 = undistortImage(I2,stereoParams.CameraParameters2);
faceDetector = vision.CascadeObjectDetector;
face1 = step(faceDetector,I1);
face2 = step(faceDetector,I2);
center1 = face1(1:2) + face1(3:4)/2;
center2 = face2(1:2) + face2(3:4)/2;
point3d = triangulate(center1, center2, stereoParams);
distanceInMeters = norm(point3d)/1000;
distanceAsString = round(distanceInMeters,2);
<-----------Training.M-------->

%This Script is used to create DataSet file db.mat
% Euclidean Distance, depth and class are calculated, class is inputted by User and then stored in db.mat
clear all;
close all force;
prompt = 'Enter Name';
N  = input(prompt);
setGlobalName(N);
for i=1:5
Depth;
tic;pause(0.01);toc;
DetectFaceandEyes;
calcEyedistance;
StoreEyeDistance;
end

----------DetectFaceandEyes.M-------

%This Script is used to detect the face by using viola jones Algorithm
%Face is Detected
 NotYet = false;
  C = webcamlist;
  cam=webcam(C{3});
  preview(cam);
  while ~NotYet
      tic;pause(0.10);toc;
      I = snapshot(cam);
      faceDetect = vision.CascadeObjectDetector();
      bbox=step(faceDetect,I);
      if ~isempty(bbox)
          NotYet = true;
          break;
      end
  end
  closepreview;
  clear('cam');
  face = imcrop(I,bbox);

----------calcEyedistance.M-------

    %Centroid of both eyes is calculated
      %RightEye 
      RighteyeDetect = vision.CascadeObjectDetector('RightEye','MergeThreshold',24);
      Righteyebox=step(RighteyeDetect,face);
      n=size(Righteyebox,1);
      e=[];
      for it=1:n
          for j=1:n
              if (j > it)
                  if ((abs(Righteyebox(j,2)-Righteyebox(it,2))<68)&& (abs(Righteyebox(j,1)-Righteyebox(it,1))>40))
                      e(1,:)=Righteyebox(it,:);
                      e(2,:)=Righteyebox(j,:);
                      d=1;
                      break;
                  end
              end
          end
          if(d == 1)
              break;
          end
      end

      Righteyebox(1,:)=e(1,:);
      Righteyebox(2,:)=e(2,:);
      c=Righteyebox(1,3)/2;
      d=Righteyebox(1,4)/2;

      %Left Eyes  Centroid
      eyeCenter1x=Righteyebox(1,1)+c+bbox(1);
      eyeCenter1y=Righteyebox(1,2)+d+bbox(2);
      e=Righteyebox(2,3)/2;
      f=Righteyebox(2,4)/2;

      %Right Eyes  Centroid
      eyeCenter2x=Righteyebox(2,1)+e+bbox(1);
      eyeCenter2y=Righteyebox(2,2)+f+bbox(2);
      euclideanDistance = CalcDistance(eyeCenter1x,eyeCenter1y,eyeCenter2x,eyeCenter2y);
      euclideanDistance = round(euclideanDistance);

----------StoreEyeDistance.M-------

%Stroing All three values in db.mat
% Taking value of class from User
  prompt = 'Enter Name\n';
  N = getGlobalName();
  if isempty(N)
  Name  = input(prompt);
  else
  Name = N;
  end
  try
      load db;
      euclideanDistance = round(euclideanDistance,1);
      F=[euclideanDistance Name distanceAsString];
      db=[db;F];
      save db.mat db 
  catch
      db =[euclideanDistance Name distanceAsString];
      save db.mat db
  end

----------EyeClassifier.M-------

% Test Face for Face Reconigastion

clc;
clear all;
close all;
Depth;
tic;pause(0.01);toc;
DetectFaceandEyes;

%%Find Out Which Class it Belong
calcEyedistance;

% Load Dataset
load db.mat
EDtrain =db(:,1);
Ntrain = db(:,2);
Dtrain = db(:,3);
j=0;

% Compare with Database
for (i=1:size(EDtrain,1))        
    if ((Dtrain(i) == distanceAsString) && (EDtrain(i)==euclideanDistance))
             j=i;
             break;
    elseif ((Dtrain(i) == distanceAsString) || (EDtrain(i)==euclideanDistance))
             j=i;
             break;
    end
end

%Checking For Class

if(j~=0)
 det_class=Ntrain(j);
 msgbox(strcat('detected class=',num2str(det_class)));
else
    msgbox('This is not trained Face');
end

% Train DataSet again if Face not matched
 if(j==0)
  StoredEyDistance;
 end

I'm Getting 3 error most the time , Please tell me How can i remove this error i learned matlab by myself and i'm weak in concept of matlab 1

Index exceeds matrix dimensions.

Error in Depth (line 24)
center2 = face2(1:2) + face2(3:4)/2;

2

Undefined function or variable 'd'.

Error in calcEyedistance (line 26)
    if(d == 1)

3

Error in imshow (line 222)
      images.internal.imageDisplayParseInputs({'Parent','Border','Reduce'},preparsed_varargin{:});
Error in imcrop>parseInputs (line 252)
            imshow(a,cm);
Error in imcrop (line 93)
[x,y,a,cm,spatial_rect,h_image,placement_cancelled] =
parseInputs(varargin{:});
Error in DetectFaceandEyes (line 19)
face = imcrop(I,bbox);
Error in Trianing (line 10)
DetectFaceandEyes;

--------CalcDistance.M------

function euclideanDistance = CalcDistance(x1, y1, x2, y2) 
euclideanDistance = sqrt((x2-x1)^2+(y2-y1)^2);

[Stalin Samuel]

What is your requirement?

[Ankit Singh]

1. Real Time application 2. Data should be stored in text format inside .mat file 3. Face Detection with person name

Suppose ankit Singh is testing the software His face will be trained first compare with stored dataset which has eye distance and depth from the camera

Answer 1

[Image Analyst]

Brett Shoelson has a face recognition demo here: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A911

Eye distance/separation doesn't seem like a really robust method. There are hundreds of people that will have the same distance, plus it will vary even on the same person depending on how close that person is to the camera.

We're not really able here to do the kind of consulting to help you build a face recognition app. It looks like it's going to take more than 5 minutes or so.

[Ankit Singh]

But i have to developed a project on this i have studied this two base paper

[Image Analyst]

OK, good luck. If you want help from us you have to ask really specific target questions about your code. We can't spend unlimited amounts of time to help you on your journey. Do you have a specific question or have an error message? "Can anyone please help me. i'm trying for 6 months" is not really specific enough, nor does it encourage us to try to tackle your project - something you have been unable to solve after 6 months of intensive trying.

[Ankit Singh]

Ok Please help me to solve 3 error

[Image Analyst]

For 1, what is the size of the face2 array? Evidently it's not as long as 4 elements. You don't even show where you calculate it. Is it size(face)?????

For 2, you need to initialize d to be zero before the loop starts, because d won't automatically have some value unless you set it.

For #3, what is I and bbox? What does this tell you

whos I
whos bbox

[Ankit Singh]

Thank You so much for the solution of D. the size of the face1 and face2 is 0X4 double when i get the error Without Error 1x4 Double for Both face1 and face2

[Image Analyst]

Well there you go. You have to check if it's 0 rows and do something, like skip to the next frame because there are no faces in this frame, or something like that.

Answer 2

[ahmed nasr]

did u know the problem with number 1 ?? am stuck with the same error ...