Fuzzy model based on RGBD images to identify biometrical facial geometry

Automatic face recognition is an important and complex task. Therefore, it is necessary to design efficient algorithms able to cope with all the different problems of face recognition. Thus, a great progress has been done using the actual vision technology. For example, a lot of topography description of the facial geometry information has not been used for the lack adequate technology. In this paper, we present an algorithm to recognize the human face, and it uses inputs from RGBD images provided by the camera and depth sensor, Kinect. The images are translated to a point cloud structure to recognize facial geometry patterns on the topography with fuzzy rules set describing the pattern algebra of each point shape identifying the connectivity pattern with their neighbors. First, it finds the nose tip which is segmented with concentric squares to evaluate symmetry. Then the face is segmented in eight different areas. The algorithm was tested in an interview scenario with a person with free range of movements.