one of the innovations that were incorporated into the Kinect SDK 1.5 was the ability to detect points of the face. Be careful, some people think that this can give us the ability to perform facial recognition, then not. What we can do is to identify more than 80 points of a face and from the same…. well what we want. A model that we could have reference is as follows:
If we are lucky to have eastern ancestry, because a face recognition points will be similar to the previous photo. If on the other hand are ugly like myself, since we will see something similar to the following:
And how facial recognition? the first thing we will see is that the code is open so anyone can fiddling with it and play with it to suit your needs.
The idea is that from the position of the Joint of the head (Head) which recognizes the Kinect sensor recognizes a virtual area around the head and Face Tracing classes return us an array with points in only 2 dimensions in order to have the references as the previous model. (there is much more to explain about how to get to the 3D of the 2D, but that goes for later).
As we can see above, there are 87 points identified in a Face. (Yes, I took the work of count them one by one) In addition to the 87 points presented in the image of our Asian friend, Face Tracking routine identifies 13 points. For example, eye Center, the center of the nose, the contours of the face and the edges of the area where the process of the face tracking is performed.
In the next post I’ll describe a step by step to show the ejemplitos of the images.
Saludos @ La Finca