Background Subtraction demo

An example using drawContours to clean up a background segmentation result.

This program demonstrates a simple method of connected components clean up of background subtraction. When the program starts, it begins learning the background. You can toggle background learning on and off using the checkbox.

Sources:

Contents

Set up video source: video file or camera

fname = fullfile(mexopencv.root(),'test','768x576.avi');
if exist(fname, 'file') ~= 2
    % download video from Github
    disp('Downloading video...')
    url = 'https://cdn.rawgit.com/opencv/opencv/3.1.0/samples/data/768x576.avi';
    urlwrite(url, fname);
end
cap = cv.VideoCapture(fname);
%cap = cv.VideoCapture(0); pause(2);
assert(cap.isOpened(), 'Cannot open video source');
cap.PosFrames = 750;

frame = cap.read();
assert(~isempty(frame), 'Cannot read data from the video source');

Create a background subtractor

bs = cv.BackgroundSubtractorMOG2('VarThreshold',10);

ZR = zeros(size(frame,1), size(frame,2), 'uint8');
niters = 3;

Prepare window

hFig = figure('Position',[100 100 840 630], ...
    'KeyPressFcn',@(o,e)setappdata(o,'flag',true));
setappdata(hFig, 'flag',false);
subplot(221), hImg(1) = imshow(frame); title('video')
subplot(222), hImg(2) = imshow(frame); title('BG model')
subplot(223), hImg(3) = imshow(ZR); title('FG segmented')
subplot(224), hImg(4) = imshow(ZR); title('FG largest component')
hCB = uicontrol('Style','checkbox', 'Position',[20 20 200 20], ...
    'String','Update Background Model', 'Value',true);

Main loop

while ishghandle(hFig)
    % get next frame
    frame = cap.read();
    if isempty(frame), break; end

    % determine whether to update BG model or not, then compute FG mask
    update_bg_model = get(hCB,'Value');
    if update_bg_model
        learnRate = -1; % automatically chosen learning rate
    else
        learnRate = 0;  % dont update BG model
    end
    fgmask = bs.apply(frame, 'LearningRate',learnRate);

    % process mask and extract largest connected component
    bw = cv.dilate(fgmask, 'Iterations',niters);
    bw = cv.erode(bw, 'Iterations',niters*2);
    bw = cv.dilate(bw, 'Iterations',niters);
    [contours, hierarchy] = cv.findContours(bw, ...
        'Mode','CComp', 'Method','Simple');
    if ~isempty(contours)
        maxArea = 0;
        largestComp = 0;
        idx = 0;
        while idx >= 0
            % compute contour area
            a = abs(cv.contourArea(contours{idx+1}));
            if a > maxArea
                maxArea = a;
                largestComp = idx;
            end
            % iterate through all the top-level contours
            idx = hierarchy{idx+1}(1);
        end
        out_frame = cv.drawContours(ZR, contours, ...
            'ContourIdx',largestComp, 'Hierarchy',hierarchy, ...
            'Color',255, 'Thickness','Filled');
    else
        out_frame = ZR;
    end

    % update images
    set(hImg(1), 'CData',frame)     % video frame
    if update_bg_model
        % get the current BG model
        set(hImg(2), 'CData',bs.getBackgroundImage())
    end
    set(hImg(3), 'CData',fgmask)    % FG mask
    set(hImg(4), 'CData',out_frame) % largest contour in FG mask

    % Terminate if any user input
    flag = getappdata(hFig, 'flag');
    if isempty(flag)||flag, break; end
    pause(0.1);
end

release camera

cap.release()