Face Detection In Video

Greeting to all, I'm now working on Viola Jones face detection which in frame by frame. But now the problem i face are is hard to storing the face segmented in a specified variables. Let's said, there're 3 faces was detected in this face detection where we named them as face 1 to face 3 and stored them into variable bbox(1,:) to bbox(3,:). but there're only 1 face remaining for next frame which was face 3. In this case, the face3 will store at bbox（1,:) instead of bbox(3,:). What if i want to empty the variable of bbox(1,:) and bbox(1,:) when face1 and face2 were disappeared. Which meant if there's another new face detected or face1 re-detect, it will direct store it into another new array of this variable. This problem already costs me for few days and is still surrounded. I'm attached my code at below and hope someone can help me to solve this ‘BIG PROBLEM' for me. Thanks to those who willing to review my question.

clear; faceDetector = vision.CascadeObjectDetector();

bbox = [100 100 100 100];

bboxx0 = [100 100 100 100];

bboxx1 = [100 100 100 100];

bbox1 = [100 100 100 100];

vidDevice = imaq.VideoDevice('winvideo', 1, 'YUY2_640x480', ... 'ROI', [1 1 640 480], ... 'ReturnedColorSpace', 'rgb' );

set(vidDevice.DeviceProperties, 'FrameRate', '30');

boxInserter = vision.ShapeInserter('BorderColor','Custom',... 'CustomBorderColor',[255 255 0]);

textInserter = vision.TextInserter('%d','LocationSource','Input port','Color',[255, 255, 0],'FontSize',12);

nFrame = 300; max =1;

vidInfo = imaqhwinfo(vidDevice);

vidHeight = vidInfo.MaxHeight;

vidWidth = vidInfo.MaxWidth;

videoPlayer = vision.VideoPlayer('Position',[300 100 640+30 480+30]);

mov(1:nFrame) = ... struct('cdata', zeros(vidHeight, vidWidth, 3, 'uint8'),... 'colormap', []);

mov1(1:nFrame) = ... struct('cdata', zeros(vidHeight, vidWidth, 3, 'uint8'),... 'colormap', []);

for k = 1:nFrame

tic;

videoFrame = step(vidDevice);

bbox = 5 * faceDetector.step(imresize(videoFrame, 1/5));

numface = size(bbox,1);

for m = 1: size(bbox,1)

if ~isempty(bbox)

bboxx0(m,1) = bbox(m,1)+bbox(m,4)*0.15;

bboxx0(m,2) = bbox(m,2)+bbox(m,4)*0.45;

bboxx0(m,3) = (bbox(m,4)/5);

bboxx0(m,4) = (bbox(m,3)/5);

bboxx1(m,1) = (bbox(m,1)+bbox(m,4)*0.6);

bboxx1(m,2) = (bbox(m,2)+bbox(m,4)*0.45);

bboxx1(m,3) = (bbox(m,4)/5) ;

bboxx1(m,4) = (bbox(m,3)/5) ;

end

end

videoOut = step(boxInserter, videoFrame, bbox);

release(boxInserter);

for m = 1:numface

strings =m;

videoOut = step(textInserter,videoOut,int8(strings),[bbox(m,1)+5 bbox(m,2)+5]);

release(textInserter);

end

for m = 1: numface

mov1(m,k).cdata = imcrop(videoOut,bboxx0(m,:)); % right

mov(m,k).cdata = imcrop(videoOut,bboxx1(m,:)); % left

end

if (size(bboxx0,1)>numface)

    for n = 0: size(bboxx0,1)-numface

    bboxx0(size(bboxx0,1)-n,:) = [];

    bboxx1(size(bboxx1,1)-n,:) = [];

    break;

    end

end

videoOut = step(boxInserter, videoOut,bboxx0);

release(boxInserter);

videoOut = step(boxInserter, videoOut, bboxx1);

step(videoPlayer, videoOut);

x1(k) = toc;

end

x = sum(x1);

release(videoPlayer);

release(vidDevice);

Answer 1

[Dima Lisin]

There is an example showing how to detect and track multiple faces in video, which may be helpful.