How can I eliminate ghost detection?

I am tracking an object but I got a problem because my program detects a ghost object. How can I eliminate this? A screenshot is shown.

[KSSV]

How we can comment/ help with out knowing code?

[rrv]

im = (abs(input_image(:,:,1)-background(:,:,1)) > threshold) | (abs(input_image(:,:,2) - background(:,:,2)) > threshold) ...
                    | (abs(input_image(:,:,3) - background(:,:,3)) > threshold);             
b = bwmorph(im,'close');
im = bwmorph(b,'open');        
im = bwmorph(im,'erode',2); 
biggest_object = bwlabel(im,8);
object = regionprops(biggest_object);
N = size(object,1);
if N < 1||isempty(object)
    return
end
s=find([object.Area]<200);
if ~isempty(s)
    object(s)=[ ];
end
N=size(object,1);
if N < 1 || isempty(object)
    return
end

for n=1:N
    hold on
    centroid = object(n).Centroid;
    C_X = centroid(1);
    C_Y = centroid(2);
end

[Walter Roberson]

The above code loops over the number of centroids but does nothing with them, and does not display anything.

It would help if we had the original image to test with. And the value of threshold .

Answer 1

[Walter Roberson]

You have attempted to index your image array as (X, Y) . That is incorrect: images arrays need to be indexed as (Y, X) in MATLAB. The first index in MATLAB, the row index, is considered to be the vertical distance (so the Y) and the second index in MATLAB, the column index, is considered to be the horizontal distance (so the X)

Answer 2

[Image Analyst]

You can get rid of half that program just by using bwareafilt():

biggest_object = bwareafilt(im, 1); % Extract largest blob only.
props = regionprops(biggest_object, 'Centroid');
if length(props) == 0  % If nothing is there, bail out.
  return;
end
% Get centroid and plot it.
xCentroid = props.Centroid(1);
yCentroid = props.Centroid(2);
hold on;
plot(xCentroid, yCentroid, 'r+', 'MarkerSize', 30);

No need for the for loop, calls to size(), and other things you have there.