Evaluation of fusion methods for latent fingerprint matchers

Matcher fusion is a recognized approach for improving biometric system performance. Component matchers may be “encapsulated,” reducing the need for understanding the inner workings of each matcher, and facilitating interchange of matchers. In large part, our interest in fusion was to determine how much performance “headroom” existed with current matcher technology. We employed five different latent fingerprint matchers. These matchers can use alternate input data (features), allowing the influence of “data type” to be investigated. Numerical results show it is possible to reduce the final candidate list to two to six candidates, with the probability that the true mate appears in the top (first) position boosted by 6-15%-points.