object detection using RGB and image

vid=videoinput('winvideo',1,'YUY2_640x480');
%Set the frames&RGB
set(vid,'FramesPerTrigger',inf);
set(vid,'ReturnedColorspace','rgb')
%Set timer of screenshoting of images per millisecond
vid.FrameGrabInterval=0.5;
%Acquiring video
start(vid);
%Set number of Frames in video
while(vid.FramesAcquired<=10000)
    %To store extracted frames
    data=getsnapshot(vid);
    %extracting the Red color from grayscale image
    diff_im=imsubtract(1.3*data(:,:,1),rgb2gray(data));
    %Filtering the noise
    diff_im=medfilt2(diff_im,[3,3]);
    %Converting grayscale image into binary image
    diff_im=im2bw(diff_im,0.18);
    %remove all pixels less than 300 pixel
    diff_im=bwareaopen(diff_im,300);
    %Draw rectangular boxes around the red object detected & label image
    bw=bwlabel(diff_im,8);
    
    stats=regionprops(bw,'BoundingBox','Centroid','Orientation');
   
    %Show image
    imshow(data);
    
    hold on 
    %create a loop for the regtangular box
    for(object=1:length(stats))
        %saving data of centroid and boundary in BB & BC 
        bb=stats(object).BoundingBox;
        bc=stats(object).Centroid;
        be=stats(object).Orientation;
        %Draw the rectangle with data BB & BC
        rectangle('Position',bb,'EdgeColor','r','LineWidth',2)
        %Plot the rectangle output
        plot(bc(1),bc(2),'-m+')
        %Output X&Y coordinates
        a=text(bc(1)+15,bc(2), strcat('X: ', num2str(round(bc(1))), '    Y: ', num2str(round(bc(2)))));
        b=text(bc(1)-50,bc(2)+45, strcat('angel:', num2str(round(be))));
    end
       hold off
end
stop(vid); 
%Flush the memory
flushdata(vid);
clear all;

I use this code to detect the red color of any object and i want to detect specific object from images using imread command but i am not perfect in using matlab to read the image then compare it with the live video to check if they are similar then start creating rectangle around the object so can anyone help me to modify the code

Answer 1

[Image Analyst]

I have code to track a colored sharpie marker. I'm attaching it. Adapt as needed.

[saeed ahmed]

But my task is to track a Coca-Cola can so I need to use image processing to measure its features. Then to detect it, for example if any other type of can, like Pepsi, is in front of camera it will skip and no action will be taken.

[Image Analyst]

Sounds like fun. Good luck.

If you need any help, see my File Exchange for more demos. Or ask a question here.

[saeed ahmed]

Ok i will check it

[saeed ahmed]

i didnt find yet my target yet do you have any idea how to make my task ?

[Image Analyst]

Maybe you can upload a short video with your target moving in it.

Answer 2

[saeed ahmed]

More explanation please

[Image Analyst]

I'm sure my code is well explained. I put way, WAY more comments into my programs than anyone else.

More data (your video) please.

[saeed ahmed]

ok but now i have extract features of my object and i want to start a video and compare the seen object by webcam to my input image that i extracted its feature , can i do that ??

[Image Analyst]

Yes, did you see the "deep learning in 9 lines of code" video that Mathworks has posted here? Search for Deep Learning in the main web site (not Answers) and I'm sure you will find it. It does that.

[saeed ahmed]

ok i will check it and if i cant reach it i will back for you , Thank you much for your time

[saeed ahmed]

now i give up searching about this comparsion because i didnt reach anything , but i will try adjust my code i want to detect the coordnates of my object place on a circle table shall i divide the circle into parts to detect the right position of the object place on this circle

[Image Analyst]

That should be pretty easy. Upload the code to process one frame, plus a frame (image) that it can work with, if you still have trouble.

You can see My Image Segmentation Tutorial if you need help.

[saeed ahmed]

%This Function is used to detect the objects from the lower plateform

%And also identifies its location and orientation measured from the right

%axis

clear all;

close all;

%connecting with the camera

vid=webcam;

%you can change the snapshot() with imread if you don't have a camera and want to %test your code

a=snapshot(vid);

%a=imread('testNow.jpg');

sum=0;

s1=floor(size(a(:,:,2),1)/2);

s2=floor(size(a(:,:,2),2)/2);

M=[s1 s1 s1 s1 s1];

K=[s2 s2 s2 s2 s2];

%Subtracting the green matrix from the rgb image

imggreen=imsubtract(1.2*a(:,:,2),rgb2gray(a));

%median filter to remove the noise if existed

imggreen=medfilt2(imggreen,[3,3]);

%transforming te image into binary

imggreen=im2bw(imggreen,0.18);

%removing areas below 300 pixels

imggreen=bwareaopen(imggreen,300);

figure,imshow(imggreen)

Spacing_cols=floor(size(imggreen,2)/6);

Spacing_rows=floor(size(imggreen,1)/6);

for(i=1:5)

sum = sum + Spacing_cols;

N = find(imggreen(:,sum));

if(size(N,1)~=0)

M(i)=N(1); % to get the numbers of rows of 1

end

end

sum=0;

for(j=1:5)

sum = sum + Spacing_rows;

L = find(imggreen(sum,:));

if(size(L,2)~=0)

K(j)=L(1); % to get the numbers of Col.s of 1

if(j==3)

v = L(size(L,2))-L(1)+2.5

end

end

end

M_Avg=floor((M(1)+M(2)+M(3)+M(4)+M(5))/5);

K_Avg=floor((K(1)+K(2)+K(3)+K(4)+K(5))/5);

Coordinates=zeros(size(imggreen,1),size(imggreen,2));

Coordinates(M_Avg,:)=1;

Coordinates(:,K_Avg)=1;

[x,y,d,RotAngle]=ObjectDetection(a,M_Avg,K_Avg,v);

figure,imshow(Coordinates | d)

hold on

text(K_Avg,10,['+ve Y-axis'],'BackgroundColor',[0 1 1]);

text(0,M_Avg,['+ve X-axis'],'BackgroundColor',[0 1 1]);

ObjectDetection

%This function is called inside the Detection.m file to locate the centroid

%and the orientation

function[x,y,A,angle]=ObjectDetection(rgbImage,x_zero,y_zero,scale)

prod_image=rgbImage;

%thresholding the image channel by channel to get the best result than...

%...thresholding the image at one time.

red=prod_image(:,:,1);

green=prod_image(:,:,2);

blue=prod_image(:,:,3);

I1=im2bw(red,0.6);

I2=im2bw(green,0.6);

I3=im2bw(blue,0.6);

%Anding the channels again

m=I1&I2&I3;

figure,imshow(m)

%Apply Median Filter to remove the noise.

A=medfilt2(m);

A=imerode(A,strel('square',30));

%remove all areas below 500 Pixels.

A=bwareaopen(A,500);

% Filling the holes in the white areas

%A=imfill(A,'holes');

figure,imshow(A)

labelled=bwlabel(A);

props=regionprops(labelled,'BoundingBox','Centroid','Orientation');

% box=props(1).BoundingBox;

% rectangle('position',box,'EdgeColor','r','LineWidth',2)

figure,imshow(prod_image)

hold on

center=props(1).Centroid;

% Plotting the Centroid

plot(center(1),center(2),'-m+')

x = (center(1) - y_zero)/(scale*10/12);

y = (center(2) - x_zero)/(scale*10/12);

angle=props(1).Orientation

text(round(center(1))-50,round(center(2))+20,['The Cenroid','(',...

num2str(round(x)),', ',num2str(round(y)),')'],'BackgroundColor',[0 1 1]);

text(round(center(1))-50,round(center(2))+45,['Rotation Angle from right horizontal',' ',num2str(angle)],'BackgroundColor',[0 1 1]);

end

can i convert this code into a live video ??

[Image Analyst]

Yes, just use either

1. getsnapshot() to read from a live video stream, or
2. read() to read from a video file, or
3. imread() to read from a still image file.

[saeed ahmed]

when i use getsnapshot it gives me error could you please add me the important command to convert it into live video stream