Abstract—This paper proposed a three-dimensional face recognition method. It is different from the conventional face recognition methods. The three-dimensional shape of the face is acquired by using a Kinect sensor, and the face is recognized by using point pair feature. Kinect sensor performs skeletal recognition with the depth image to estimate the face region. A robust face recognition becomes possible by the method of using point pair feature. To confirm the effectiveness of the proposal method, at first, an experiment was conducted to determine a threshold for face identification. By the experimental result, a value matching 70% of the 3D point cloud of the face area was set as the threshold for face identification. Next, we registered the face model of 10 subjects (9 males, 1 female), and conducted an authentication experiment based on the determined threshold. As a result of the experiment, the authentication rate of 100% was obtained without an others acceptance and a principal denial (false rejection).

Index Terms—face recognition, image processing, point pair feature, three-dimension, 3D point cloud

Cite: Yoshihiro Sato and Yue Bao, "3D Face Recognition without Using the Positional Relation of Facial Elements," Journal of Image and Graphics, Vol. 6, No. 1, pp. 33-38, June 2018. doi: 10.18178/joig.6.1.33-38