Modeling biometric template update with Ant Colony Optimization

In this paper we present a novel template update algorithm based on the semi-supervised learning algorithm Aggregation Pheromone Density Based Semi-Supervised Classification (APSSC). APSSC is an Ant Colony Optimization (ACO) based algorithm, and it is inspired by the social behaviour of ants. The automatic update of biometric templates is modeled by representing stored data as ants, grouped into two colonies. One colony is populated by the ants representing the enrolled template related to a given client. The second colony is populated by the ants representing the data used as impostor training. The biometric template update process is modeled as the aggregation of ants to colonies. We tested the APSSC algorithm on the BANCA 2D faces dataset. To the extent of our knowledge, this is the first time that such methodology has been proposed for template update. In our experiments we show that a modified version of APSSC could be a promising algorithm to deal with biometrics template update.