, , , , Michael Chumakov
Background: Stability is a required definitional property for a biometric to be useful. Quantitative statements of stability are operationally important as they dictate re-enrollment schedules e.g. of a face on a passport. Ophthalmologists consider the iris to be stable, and accordingly iris recognition accuracy was thought to be invariant over time[26, 16]. This view held until several recent empirical studies suggested otherwise. Two of these, using separate iris image collections from the University of Notre Dame, reported a large template-ageing effect[5, 24]. The studies claimed to have excluded several possible causes of the observed ageing, but could not conclude that the iris texture itself was changing. Their results, however, were widely publicized[53, 19, 3] with statements such as irises, rather being stable over a lifetime, are susceptible to ageing effects that steadily change the appearance over time. A further study, however, noted pupil-dilation as the primary causal variable. Detection of long-term ageing trends is complicated by short-term stochastic variations inherent in acquiring digital images from a live analog anatomical source - see Figure1. Approach: We quantify time variation in iris recognition accuracy in two ways. First we produce rate-of-change estimates for up to 122,000 frequent travelers using a fixed iris recognition system for up to 9 years. Second, we apply iris recognition algorithms to the images of 217 individuals used in the Notre Dame studies. The algorithms produce pupil dilation and exposed iris area measures which we relate to recognition outcomes. Additionally, we review other published studies, formulate recommendations for conduct of biometric ageing studies and for the mitigation of ageing effects in operational systems.
NIST Interagency/Internal Report (NISTIR) - 7948