Relaxing the constraints on image capture for iris recognition systems

Iris recognition is considered to be one of the most accurate biometrics, but user inconvenience during the image acquisition phase has limited its widespread use. Image capture is largely constrained to well-controlled situations, where subjects must remain relatively stationary or within a capture "volume" close to the camera. As a consequence, iris recognition systems have a reputation for being borderline intrusive, and less friendly for both subjects and operators. To support the development of a more natural and acceptable iris capture system, we have sought to develop a pre-processor driven imaging system that predicts a maximal opportunity window for iris capture for a subject engaged in natural motion based on predictive head and eye movement algorithms. This paper describes a first-generation prototype iris capture system that utilizes this approach. A wide field of view camera is used to track a person's face and provide head pose data as input to the predictive algorithm. The algorithm is then used to direct a second narrow field of view camera to capture the iris image more reliably. This system serves as a platform for further development of head movement prediction algorithms used to enhance the probability of iris capture in moving or uncooperative subjects.