How to crop images and save it to matrix after vision.For​egroundDet​ector

Hello. Vision.ForegroundDetector is so amazing!

Could you let me know how to get several images inbox as a matrix?

I want to save these images as matrix.

I want to use several images like this. How can I crop this like image?

I am using this code.

videoSource = VideoReader('shahar_run.avi');
detector = vision.ForegroundDetector(...
       'NumTrainingFrames', 5, ...
       'InitialVariance', 30*30);
blob = vision.BlobAnalysis(...
       'CentroidOutputPort', false, 'AreaOutputPort', false, ...
       'BoundingBoxOutputPort', true, ...
       'MinimumBlobAreaSource', 'Property', 'MinimumBlobArea', 250);
shapeInserter = vision.ShapeInserter('BorderColor','White');
videoPlayer = vision.VideoPlayer();
while hasFrame(videoSource)
     frame  = readFrame(videoSource);
     fgMask = detector(frame);
     bbox   = blob(fgMask);
     out    = shapeInserter(fgMask,bbox);
     videoPlayer(out);
     pause(0.1);
end

[Ameer Hamza]

Can you attach the actual image or the video file?

[Kong]

I attached the video. Thank you so much.

Answer 1

[Ameer Hamza]

Check the following code

videoSource = VideoReader('denis_wave2.avi');
detector = vision.ForegroundDetector(...
    'NumTrainingFrames', 5, ...
    'InitialVariance', 30*30);
blob = vision.BlobAnalysis(...
    'CentroidOutputPort', false, 'AreaOutputPort', false, ...
    'BoundingBoxOutputPort', true, ...
    'MinimumBlobAreaSource', 'Property', 'MinimumBlobArea', 50);
shapeInserter = vision.ShapeInserter('BorderColor','White');
videoPlayer = vision.VideoPlayer();
bounding_boxes = {};
while hasFrame(videoSource)
    frame  = readFrame(videoSource);
    fgMask = detector(frame);
    bbox   = blob(fgMask);
    for i=1:size(bbox, 1)
        x = bbox(i, :);
        bounding_boxes{end+1} = fgMask(x(2):sum(x([2 4])), x(1):sum(x([1 3])));
    end
    out    = shapeInserter(fgMask,bbox);
    videoPlayer(out);
    pause(0.1);
end

The bounding_boxes is a cell array and its each element is a cropped image. You can check the output as

imshow(bounding_boxes{1})
imshow(bounding_boxes{2})

[Kong]

Thank you so much.

Could you explain how to save several bounding_boxes as a matrix in the video?

I want to use the matrix for next step like classification.

[Ameer Hamza]

Kong, this code save all the bounding boxes in the cell array bounding_boxes. The basic concept is this: each row in bbox represent a represent [x y width height] around the blob. You just need to use correct indexing to extract that blob from the image. For exaple

fgMask(x(2):sum(x([2 4])), x(1):sum(x([1 3])))

is equivalent to

fgMask(y:y+height, x:x+width)

[Kong]

Thanks a lot.

When I used this code, I got 8 bounding_boxes images.

Could I get an idea to obtain good shape of human motion for classification?

Each image is not good for classification.

figure;
subplot(181); imshow( bounding_boxes{1} );
subplot(182); imshow( bounding_boxes{2} );
subplot(183); imshow( bounding_boxes{3} );
subplot(184); imshow( bounding_boxes{4} );
subplot(185); imshow( bounding_boxes{5} );
subplot(186); imshow( bounding_boxes{6} );
subplot(187); imshow( bounding_boxes{7} );
subplot(188); imshow( bounding_boxes{8} );

[Ameer Hamza]

I am not sure how to do that. It requries a fairly domain specific knowledge. Maybe try to start a new question so that someone with the required expertise can answer or trying posting such a question on a forum related to computer vision.

[Kong]

Thank you so much. I will.

[Image Analyst]

Get state of the art algorithms in gesture recognition here in Vision Bibliography