A facial recognition method based on 3-D images analysis for intuitive human-system interaction

Facial recognition technology is an important part of human-system interaction, which can be used in ID verification, man-machine interface, etc. Many methods based on the analysis of a color image have been proposed. However, the real-time performance or accuracy of the existing recognition approaches is not satisfying, which can be influenced by factors such as lighting environment, position change, face make-up, etc. In this paper, we propose a facial recognition method based on three-dimensional (3-D) images measurement and 3-D image histogram analysis. We use the optimal intensity-modulation pattern projection (OIMP) technique to obtain the 3-D information of face. Since the OIMP technique needs only one single projection, it is easy to apply to human face measurement. The recognition procedure of the method we proposed is as follow. First, set up a 3-D database to store facial information obtained by using OIMP measurement technique. Only the front face images are needed. Second, capture the 3-D face image of the identified person. Then do the recognition by comparing histogram of saved 3-D image in database with the captured 3-D image. If the degree of histograms´ similarity of the two 3-D images is the maximum and the value is higher than a certain threshold, then the user can be determined to be the right person. To verify the feasibility of the method, we have a group of testers including male and female of different ages and conduct a verification experiment. In the different lighting environments and different angles for faces, the average recognition accuracy was higher than 90%.