In our recent article, “AFIS – what makes a great solution?”, we looked at what factors are necessary to succeed in the competitive AFIS market. Here, in part 2, we’ll take a closer look at why speed is so important in large-scale AFIS solutions.
Matching speed is a key factor, especially for large-scale projects with multiple millions of enrollees such as:
- Voter registration and identification for elections
- Electronic identification documents, drivers licenses, and passports
- Identity verification for banks and financial institutions
- Crime scene forensics and criminal identification
All of the above have their own specific, time-critical processing steps, running behind the scenes in real-time in order to deliver the required outputs. Fingerprint recognition speed depends on several processing stages, which are bound to biometric systems.
During the data capture stage, ultra-fast, real-time quality evaluation is what most affects not only the user experience but also the achieved quality of the biometric sample. After the capture, the final quality is calculated and displayed almost immediately, and template extraction and data formatting is performed.
In a typical AFIS, the data is then sent for pre-processing, segmentation, sequence check, further QA and template extraction, then queued for further AFIS processing. Every operation needs to be optimized in order to facilitate extremely high-volume transactions involving hundreds of millions of records.
In order to meet these demands, the matching algorithm must be capable of carrying out these operations not only on large server clusters, but also on mobile workstations, handheld terminals, and various embedded devices.
Innovatrics’ response to these challenges:
QA, Template Extraction and 1:1 Verification Speed
- Highest matching speed and top overall processing time (template extraction and verification) among accuracy leaders in NIST PFT II
- Top extraction and matching speed in NIST MINEX III
- Fastest average enrollment per finger in NIST FpVTE 2012
Identification and Deduplication Processing