Iris Recognition at a Distance
We describe experiments demonstrating the feasibility of human iris recognition at up to 10 m distance between subject and camera. The iris images of 250 subjects were captured with a telescope and infrared camera, while varying distance, capture angle, environmental lighting, and eyewear. Automatic iris localization and registration algorithms, in conjunction with a local correlation based matcher, were used to obtain a similarity score between gallery and probe images. Both the area under the receiver operating characteristic (ROC) curve and the Fisher Linear Discriminant were used to measure the distance between authentic and imposter distributions. Among variables studied, database wide experiments reveal no performance degradation with distance, and minor performance degradation with, in order of increasing effect, time (one month), capture angle, and eyewear.
Unable to display preview. Download preview PDF.
- 3.Mansfield, T., Kelly, G., Chandler, C., Kane, J.: Biometric Product Testing Final Report, Technical Report, Centre for Mathematics and Scientific Computing, National Physics Laboratory, Teddington, Middlesex, UK (2001)Google Scholar
- 5.Camus, T., Wildes, R.: Reliable and fast eye finding in close-up images. In: Proc. Int. Conf. on Pattern Recognition, pp. 389–394 (2002)Google Scholar
- 7.Bergen, J.R., Anandan, P., Hanna, K., Hingorani, R.: Hierarchical model-based motion estimation. In: Sandini, G. (ed.) ECCV 1992. LNCS, vol. 588, pp. 237–252. Springer, Heidelberg (1992)Google Scholar
- 9.Doddington, G., Liggett, W., Martin, A., Pryzybocki, Reynolds, D.: Sheep, goats, lambs and wolves: a statistical analysis of speaker performance in the NIST 1998 speaker recognition evaluation. In: Proc. of 5th Int. Conf. of Spoken Language Processing (1998)Google Scholar