Sir could i trouble you in asking for your help to look into my code and give a reason to why i dont get my needed output?

I'm doing my final year project on MOVING OBJECT DETECTION and with all the help from the internet finally finished my code and i jst wont get the Moving Object output. The video I'm feeding is a live video through my webcam.

function realVideo() NumberFrameDisplayPerSecond=10; hFigure=figure(1); source= videoinput('winvideo', 1); set(source,'FramesPerTrigger',1); set(source,'TriggerRepeat',Inf); set(source,'ReturnedColorSpace','RGB'); triggerconfig(source, 'Manual'); Timer=timer('TimerFcn', {@FrameRDisplay,source},'Period',1/NumberFrameDisplayPerSecond,'ExecutionMode','fixedRate','BusyMode','drop'); start(source); start(Timer); uiwait(hFigure); stop(Timer); delete(Timer); stop(source); delete(source); clear functions; end

% This function is called by the timer to display one frame of the figure

function FrameRDisplay(obj, event,source) persistent IM; persistent handlesRaw; persistent handlesPlot;

trigger(source); IM=getdata(source);

if isempty(handlesRaw)

   subplot(2,1,1);
  imagesc(IM);
   title('REAL TIME IMAGE');

   for i = 2:3:(length(source)-1)

    fr1 = source(i-1).cdata;       % read in frame-i-1
    fr_bw1 = rgb2gray(fr1);       % convert frame-i-1 to grayscale
    fr2 = source(i).cdata;         % read in frame-i
    fr_bw2 = rgb2gray(fr2);       % convert frame-i to grayscale
    fr3 = source(i+1).cdata;       % read in frame-i+1
    fr_bw3 = rgb2gray(fr3);       % convert frame-i+1 to grayscale

    fr_diff1 = abs(double(fr_bw1) - double(fr_bw2));  % First frame Difference cast operands as double to avoid negative overflow
    fr_diff2 = abs(double(fr_bw2) - double(fr_bw3));  % Second frame difference cast operands as double to avoid negative overflow

        for j=1:width                 % if fr_diff > thresh pixel in foreground
            for k=1:height
                if ((fr_diff1(k,j) > thresh) && (fr_diff2(k,j) > thresh))
                f(k,j) = 255;
                else
                f(k,j) = 0;
                end
            end

        end
    subplot(2,1,2);
   handlesPlot=imshow(uint8(f));
   title('MOVING OBJECT');
   end
end
end

[Image Analyst]

I don't see where you set "thresh" to any value. What kind of output do you get? What does (the badly named) binary image "f" look like? If you display fr1, fr2, and fr3, so they look as expected?

[Sean de Wolski]

So it looks like you copied the code from here:

http://www.matlabtips.com/realtime-processing/

I highly recommend citing others' work when you choose to copy it.

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup

[Sanjeev]

Image Analyst i don't get anything other than the real time video output of myself sitting in front of my pc. nothing happens when i try to display fr1,fr2,fr3. Sean de Wolski Sorry for that.

[Walter Roberson]

It doesn't even open a plot window for you?

[Sanjeev]

it opens up a blank window.. nothing else.

[Walter Roberson]

Consider how much easier it is for volunteers to understand your code when it is nicely formatted. http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup

[Sanjeev]

sorry abt that walter. Hope this is better.

function realVideo()

% Define frame rate
NumberFrameDisplayPerSecond=10;

% Open figure
hFigure=figure(1);
thresh=15;

% Set-up webcam video input
   source= videoinput('winvideo', 1);

% Set parameters for video

     set(source,'FramesPerTrigger',1);

% Go on forever until stopped
    set(source,'TriggerRepeat',Inf);

% Get a grayscale image
    set(source,'ReturnedColorSpace','RGB');
    triggerconfig(source, 'Manual');

% set up timer object
    TimerData=timer('TimerFcn', {@FrameRDisplay,source},'Period',1/NumberFrameDisplayPerSecond,'ExecutionMode','fixedRate','BusyMode','drop');

% Start video and timer object
    start(source);
    start(TimerData);

% We go on until the figure is closed
    uiwait(hFigure);

% Clean up everything
    stop(TimerData);
    delete(TimerData);
    stop(source);
    delete(source);
% clear persistent variables
    clear functions;
end

function FrameRDisplay(obj, event,source)
persistent IM;
persistent handlesRaw;
persistent handlesPlot;

trigger(source);
IM=getdata(source);

if isempty(handlesRaw)
   % if first execution, we create the figure objects
  subplot(1,2,1);
  imagesc(IM);
   title('REAL TIME IMAGE');

     % Plot first value

     for i = 2:3:(length(source)-1)

      fr1 = source(i-1).cdata;       % read in frame-i-1
      fr_bw1 = rgb2gray(fr1);       % convert frame-i-1 to grayscale
      fr2 = source(i).cdata;         % read in frame-i
      fr_bw2 = rgb2gray(fr2);       % convert frame-i to grayscale
      fr3 = source(i+1).cdata;       % read in frame-i+1
      fr_bw3 = rgb2gray(fr3);       % convert frame-i+1 to grayscale

      fr_diff1 = abs(double(fr_bw1) - double(fr_bw2));  % First frame Difference cast operands as double to avoid negative overflow
      fr_diff2 = abs(double(fr_bw2) - double(fr_bw3));  % Second frame difference cast operands as double to avoid negative overflow

          for j=1:width                 % if fr_diff > thresh pixel in foreground
              for k=1:height
                  if ((fr_diff1(k,j) > thresh) && (fr_diff2(k,j) > thresh))
                  f(k,j) = 255;
                  else
                  f(k,j) = 0;
                  end
              end
          subplot(1,2,2);
     handlesPlot=imshow(uint8(f));
     title('MOVING OBJECT')
          end

     end
  end
  end

[Walter Roberson]

What purpose does handlesRaw serve in the second routine? It is a persistent variable and some code is only done if it is empty, which would be the case if it had never been initialized. That kind of isempty() test is normally only used to determine whether some initialization needs to be done the first time through a routine, with the initialization being skipped on other runs through the routine, but in this case nothing is done if the object is already initialized, and the code that is run when it is not initializes never initializes it. This is confusing for readers: if something needs to be done every time the routine is called, then just don't protect it with that kind of test.

Anyhow, from the point in the code "Plot first value", the variable "source" gets used in many questionable ways. The variable has been passed in from the realVideo routine, and represents a videoreader object. Arrays of videoreader objects are allowed according to the documentation for disp() applied to videoreader objects, so subscripting of a videoreader object would select a single videoreader object, and length() applied to a possible array of videoreader objects would return 1 if there is only a single videoreader object in the array, such as is the case here. So the

for i = 2:3:(length(source)-1)

is going to be effectively

for i = 2:3:(1-1)

which is never going to execute.

Inside that "for" block, the videoreader object "source" has some information accessed as source(i).cdata . That would be fine if videoreader objects had a cdata method, but as far as I can tell they do not. The information that the section of code is attempting to extract is information that would have been extracted into the 4-dimensional array

IM=getdata(source);

where the 4th dimension is the frame number.

It appears that the whole "for i" loop needs to be rewritten to pull information from "IM" instead from "source"

Now there is a problem, that the code in that "for" loop expects to work through bunches of 3 frames at a time, but the videoreader object has been set to acquire 1 frame per trigger. 3 getdata() would be needed to retrieve 3 frames with that videoreader object configuration, but there is only one getdata() call in the FrameRDisplay call -- and the time interval for the timer has been set to run the fraemRDisplay routine once for every frame expected by the frame rate, so there should not be multiple getdata() calls in the FrameRDisplay routine. So you are going to need to rewrite that routine to save data in-between calls, and you are going to have to consider whether you want to apply the movement detection in non-overlapping groups of 3 frames or in overlapping groups of 3 frames.

My summary is that the code you have downloaded is too broken for use as-is and should be mostly thrown away and re-written, at least the display routine.

[Sanjeev]

thanks for your help Walter.. helped a lot! could you suggest a way to find the "length(source)"? i was thinking whether its possible to keep the camera on and get my required output till i close the program or stop its execution?

[Walter Roberson]

"source" is the video object. length(source) is 1.

It is like having a clock and asking how many clocks you have (one!) when what you should have done is looked at the clock instead of counting how many there are. getdata(source) is the way to read the video instead of counting how many cameras you have.

[Jerome Lecoq]

Hello Walter,

I am happy to modify my code at http://www.matlabtips.com/realtime-processing/ if you believe it is broken in any way. I never expected so many people to read it when I initially did it and I don't want anyone to rely on broken things.

The original code is somehow different from the version that Sanjeev made.