Improvement to the quality of recognition and availability (SpeedCluster) of Griaule’s AFIS

• Grant number: 03/07972-6
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: June 01, 2004
• End date: May 31, 2007
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Iron Calil Daher
• Grantee: Iron Calil Daher
• Company: Griaule Biometrics Ltda
• City: Campinas

Abstract

Griaule developed and commercialized a Brazilian system for the identification of fingerprints (Automated Fingerprint Identification System, AFIS). The software is in use in the states of São Paulo, Rio Grande do Norte, Sergipe and Tocantins. Tests conducted internally, based on the procedures of the University of Bologna (Unibo), demonstrate that, out of 28 international competitors, the product lies in tenth position. The first objective of this project presented here is to improve the system’s recognition rate. Adopting the EER (Equal Error Rate) metrics, widely accepted, and taking as a reference Unibo’s ranking, the intention is to achieve, in relation to that rate, an EER lower than 0.9 per cent. The second objective is to increase the availability of the Griaule AFIS using the cluster concept, guaranteeing the system’s continuity in case of the failure of one of its servers. The improvement in recognition involves the pre-processing of the digital images, the treatment of distortion and the segmentation of the images and the use of morphological information, among other topics. As for the increase in availability, the development of a search queuing scheme, fault tolerance and load distribution is envisaged.The expected result is a piece of AFIS software suited to the Brazilian reality, at a low cost and of high quality. (AU)