Multi-target planar tracking

Example of using features2d framework for multiple planar targets tracking in a video using homography matching. This sample uses the PlaneTracker class.

Sample video: http://www.youtube.com/watch?v=pzVbhxx6aog.

Sources:

% video file, and multiple targets to track [x,y,w,h]
vid = fullfile(mexopencv.root(), 'test', 'blais.mp4');
assert(exist(vid, 'file') == 2, 'Missing video file');
win = [
    136 0 366 433;   % book
    135 165 285 175  % face
];
N = size(win,1);

% open video feed, and get firs frame
cap = cv.VideoCapture(vid);
pause(1);
assert(cap.isOpened(), 'Failed to open video');
frame = cap.read();
assert(~isempty(frame), 'Failed to read frames');

% create and initialize tracker
tracker = PlaneTracker();
for i=1:N
    tracker.addTarget(frame, win(i,:));
end

% prepare plot
hImg = imshow(frame);
clr = lines(N);
for i=1:N
    % keypoints and bounding box for each tracked target
    hPts(i) = line(NaN, NaN, 'Color',clr(i,:), 'LineStyle','none', 'Marker','o');
    hLin(i) = line(NaN, NaN, 'Color',clr(i,:), 'LineWidth',3);
end

% main loop
while ishghandle(hImg)
    % read new frame
    frame = cap.read();
    if isempty(frame), break; end

    % track and update keypoints/boundaries of matched targets
    tracked = tracker.track(frame);
    for i=1:numel(tracked)
        tr = tracked(i);
        set(hPts(tr.index), 'XData',tr.pt1(:,1), 'YData',tr.pt1(:,2));
        set(hLin(tr.index), ...
            'XData',tr.quad([1:end 1],1), 'YData',tr.quad([1:end 1],2));
    end

    % update tracked objects which were not found in current frame
    idx = setdiff(1:N, [tracked.index]);
    if ~isempty(idx)
        set(hPts(idx), 'XData',NaN, 'YData',NaN);
        set(hLin(idx), 'XData',NaN, 'YData',NaN);
    end

    % display result
    set(hImg, 'CData',frame);
    drawnow;
end
cap.release();