Face recognition with Zernike moments

摘要

In recent years, considerable research has been performed on technology for verifying individual identity using human biometrics. In particular, verification technology using facial images has been extensively researched. This is because special input devices are not needed, and there is little of the psychological resistance that is found with fingerprint input. In this paper, the author proposes a method for recognition using the subspace method, and using the Zernike moment, a moment that is invariable during rotation, as a feature vector. The validity of the method is then shown. In addition, the author confirms through experiments the relationship between the circle width of the circular Zernike moment and the recognition rate, as well as between the order of the Zernike moment and the recognition rate. Furthermore, the recognition capacity of each circle is described, and experiments are performed on variations in the recognition rate when the circle with the lowest recognition capacity is eliminated. Finally, experiments comparing the author's approach to other methods are performed. The author shows that equivalent recognition results can be obtained using the Zernike moment as a feature vector whose order is far lower than other feature vectors.