Robust 3D Face Recognition
Janez Kri, Simon Dobri · 2012
Face recognition in uncontrolled environments is hindered by variations in illumination, pose, expression and occlusions of faces. Many practical face-recognition systems are affected by these variations. One way to increase the robustness to illumination and pose variations is to use 3D facial images. In this paper 3D face-recognition systems are presented. Their structure and operation are described. The robustness of such systems to variations in uncontrolled environments is emphasized. We present some preliminary results of a system developed in our laboratory. Systems for the biometric recognition of individuals assess the identity of these individuals on the basis of their physiological or behavioral characteristics, like fingerprint, face, speech, gait and iris patterns. The scope of these systems includes various kinds of access control (border crossing, access to personal information), foren- sics, law enforcement and others. Face-recognition sys- tems are some of the most popular among all biometric systems. This is mostly due to their non-intrusive nature, as the data acquisition can be performed from a distance, even without the subject's cooperation. Special attention in such biometric systems is being paid to developing the so-called smart surveillance technologies, especially to developing portals for the automatic control of border crossings (1). Although people recognize faces without any spe- cial effort, automatic face recognition with a computer presents a considerable difficulty and challenge, espe- cially if the images are acquired in an uncontrolled environment. The main factors that affect the accuracy of face-recognition systems are the variability in the illumination and the pose of the faces, expressions, occlusions (scarf, beard, glasses), time delay (signs of aging), makeup and similar occurrences during the image-acquisition task. The presence of these factors in the facial images can lead to a diminished recognition reliability.