C ++（和openCV）：在向量中累积许多Mat <Mat>

Question

I'm grabbing frame from webcam am displaying them on a window. To do a temporal analysis, I keeping the frames on a vector as I grab them. In order to test it, when the size of my vector reaches 100, I attempt to visualize all the frames that I stored so far.

The strange thing is that, indeed the vector has 100 frames inside, but they are all the same and correspond to the last frame that was captured.

Code:

Mat frame;       

    _cap.start(0);  //VideoCapture object   

    vector<Mat> seq;
    while(1)
    {
        _cap.getFrame(frame);           

        seq.push_back(frame);

        imshow("a", seq[0]);

        if (seq.size() == 100)
        {
            for (int n = 0; n < seq.size(); n++)
            {
                cout << "I'm in" << endl;
                imshow("b", seq[n]);
                waitKey(0);
            }
        }

            waitKey(30);   
    }

Answer 1

the Mat's you get from the VideoCapture all point to the same driver memory.

Mat a=b; does a shallow copy ( same pixel pointer )

so you need :

 seq.push_back(frame.clone());

here ( ie a deep copy )

Answer 2

Frame will still point to the same data field. Hence whenever you grab a new image all previous captured images will also update.

Solution is to release the frame after you have put it into the vector by calling frame.release()