We conduct research in the field of tracking faces in real time. The well-known method for the detection of persons used in the OpenCV library is too slow. However, most scientific publications refer to the method of selection of persons on the color mask. After spending binarization image complexion, you can get the field that you can then select it, and then check for compliance with the human face. Below is an example of research, designed in the form of video.
As can be seen from the video, this method allows us not only to identify the face (and rather quickly), but also to talk about the approximate angle of rotation. To eliminate false positives need to detect elements of the face, such as the eyebrows, eyes, nose, mouth. Below are screenshots of the test method for determining the eyebrows and mouth.
At present, research is ongoing.