edit -door detection

I have a code for detecting the door,but id detect only some images ,it detects only doors brown in colour,if it is other colours,it os not detecting,please help

clc
clear all

[filename, pathname] = ...
     uigetfile({'*.jpg';'*.bmp';'*.png';'*.*'},'File Selector');

   i=imread([pathname,filename]);
   I=i;
   if length(size(i)) == 3
    im =  double(i(:,:,2));
  else
    im = double(i);
  end
  cs = fast_corner_detect_9(im, 60)
  size(im)
  image(im/6)
  axis image
  colormap(gray)
  hold on
  axis off

plot(cs(:,1), cs(:,2), 'r.'),title('Corner Detection')

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%GUASSIAN SMOOTHING %%%%%%%%%%%%%%%%%%%%%%%%%%%%
G = fspecial('gaussian',[5 5],2);

Ig = imfilter(im,G,'same');

figure('name','Guassian Smoothing','numbertitle','off'),imshow(uint8(Ig))

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%RESIZING AN IMAGE%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 I1=imresize(Ig,.5)

figure('name','Reduced Resolution Image','numbertitle','off'),imshow(uint8(I1))
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%EDGE DETECTION%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

edge_detect=edge(im,'sobel');

figure('name','Edge detected Image','numbertitle','off'),imshow(edge_detect);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%DETCTION%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
cform = makecform('srgb2lab');
J = applycform(I,cform);
%figure;imshow(J);
K=J(:,:,2);
%figure;imshow(K);
L=graythresh(J(:,:,2));
BW1=im2bw(J(:,:,2),L);
%figure;imshow(BW1);
M=graythresh(J(:,:,3));
%figure;imshow(J(:,:,3));
BW2=im2bw(J(:,:,3),M);
%figure;imshow(BW2);
O=BW1.*BW2;
% Bounding box
P=bwlabel(O,8);
BB=regionprops(P,'Boundingbox');
BB1=struct2cell(BB);
BB2=cell2mat(BB1);

[s1 s2]=size(BB2);
mx=0;
for k=3:4:s2-1
    p=BB2(1,k)*BB2(1,k+1);
    if p>mx & (BB2(1,k)/BB2(1,k+1))<1.8
        mx=p;
        j=k;
    end
end
figure,imshow(I);
hold on;

rectangle('Position',[BB2(1,j-2),BB2(1,j-1),BB2(1,j),BB2(1,j+1)],'EdgeColor','g' )
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%END OF CODING%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

[Walter Roberson]

Duplicate is at http://www.mathworks.com/matlabcentral/answers/28416-reducing-the-size-of-image

Another branch of the same duplicate is at http://www.mathworks.com/matlabcentral/answers/26731-detecting-the-corners

[kash]

yes walter but i did not get answer fro any

[kash]

this code detects only some doors

[Pria Bertopeng]

iam trying to run door detection code from doordetection in Octave. And there is error

warning: the 'makecform' function belongs to the image package from Octave Forge but has not yet been implemented.

Please read http://www.octave.org/missing.html to learn how you can contribute missing functionality. error: 'makecform' undefined near line 42 column 9 error: called from doordetection at line 42 column 7

how to fix that?

thx

[Walter Roberson]

Pria Bertopeng:

The obvious fix is to use MATLAB instead of Octave.

Questions specific to Octave should be addressed to an Octave resource.

[Pria Bertopeng]

Can someone explain how this code works? especially on detection section. Thx

[Walter Roberson]

It converts to L*a*b color space, and does a thresholding on the L (brightness) channel to produce a binary image in which 0 corresponds to darker spots and 1 corresponds to brighter spots.

You could also do much the same with rgb2gray, or with rgb2hsv and taking the v channel as the brightness.

Answer 1

[Winnie]

hi...Did you write this code by yourself? Because my friend has the exact same code as you!

Answer 2

[Image Analyst]

I have no idea how challenging this is. Perhaps you have images with doors that have a known and contrasting color from the rest of the items in the image. If so, then you could use the color detection methods in my File Exchange and then check on what their area is (large = closed door, small = open door). http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862 This would be one of the easier cases.

If you're hoping for some code that will tell you "open" or "closed" for every possible door that a search of "door" images on Google might bring up, then that's way beyond your capabilities now. Some people in the "CBIR" community are working on that but it's very very challenging. For example, show me, based on only the image pixels and no metadata, images that have pictures of babies in them. Or dogs, or doors, or cars, or whatever. Look up CBIR in a web search.

[Walter Roberson]

For example there are a lot of windows that are also doors.

[Image Analyst]

Not to mention things like gates. Is a half height solid wooden gate a door? How about a full height, wrought iron, non-solid gate? Is it supposed to find the door on a car, and the door to the batteries on my electronic device, as well as doors to buildings? What if the door is glass, like in many modern buildings? I wouldn't tackle this problem unless it was very very constrained to certain types of doors.

[Walter Roberson]

Yes for the solid wooden gates: consider the traditional "Old West" saloon doors.

Answer 3

[Tanim Ahsan]

@kash Did u able to solve your problem?