Data Collection

Abstract

“Collected biometric samples or features are properly referred to as a “corpus”. The information about those images and the clients who produced them is referred to as the “data base”. Data collection, which affects both the corpus and the data base, is most relevant for machine readable travel documents (MRTD). Both the corpus and the database can be corrupted by human error [and/or security lapses] during the collection, processing, and storage process (see Ch. 14). “In fact, error rates in the collection process may easily exceed those of the biometric device. Extreme care must thus be taken during data collection to avoid both corpus (images improperly acquired) and data base (mislabeled image) errors” (Mansfield et al., 200I). There are two basic remedies for such errors. The face recognition engine should flush out such errors and/or use manual inspection. There are two types of failures that can be traced to image acquisition errors. The failure to enroll rate is the expected proportion of the population for whom the system is unable to generate repeatable [signatures] templates, while the failure to acquire rate is defined as the expected proportion of transactions for which the system is unable to capture or locate an image or signal of sufficient quality. The two errors related to image acquisition are handled by repeated attempts until success is observed. FACE IT from Visionics (now part of Viisage), during testing and evaluation, collects images over a period of I0 seconds and provides the best (pair wise) match obtained.