Identity AlgorithmsIn less than three years, Retica has become one of the top tier software and device solution suppliers. In addition to a set of iris acquisition tools that facilitate our range of iris capture devices, our software team has developed a set of core iris algorithms. Throughout the development process our goal has been to build algorithms that are accurate, fast and robust. AccuracyThe full potential of the iris biometric is embraced by highly accurate iris localization and optimized encoder design. Retica’s random-bit binary iris encoding methods have demonstrated higher levels of information content than competing methods. Large databases require very small false match rates. Retica’s Iris Encoder/Matcher tools have been subjected to independent third-party testing using unseen datasets of iris images. Tests included over 90 billion imposter matches. Our true-match-rates were shown to be stable (fell < 1.5%) at false-accept rates ranging from 1 in 1 million to 1 in 10 billion. Retica is also committed to multimodal biometric fusion. Dual iris fusion strategies have been demonstrated to reduce single-eye errors rates by an order of magnitude. High accuracy rates have been demonstrated: 99.96% TAR @ 10-4 FAR with dual iris fusion on the NIST-ICE dataset. SpeedMajor components of Retica’s iris analysis algorithms are applied at frame rate during iris acquisition. These include image quality assessment and iris localization (segmentation). A binary encoding implementation also facilitates simple, rapid iris matching methods. Matching speeds of the order of 240,000 matches per second per CPU-core for each rotation angle at full code resolution are expandable up to 3,000,000 matches per second per CPU-core with hierarchical multi-resolution code matching. RobustnessIt is important for leading iris analysis tools to support a range of different iris optical systems and system architectures. All of Retica’s software methods have been developed for use with a large range of iris image characteristics. Algorithms have been designed to mitigate poor iris image quality found in publicly available iris image datasets. In addition, they have been developed to work with the full range of our proprietary optical engines (standoff distances ranging from 0.05 to 50 meters). Finally, our software tools can be applied on a variety of platforms via their adoption of required interoperability standards, e.g. EBTS, ANSI-INSITS, and ISO/IEC compliance. Retica System’s core fusion identity algorithmsPushing the boundaries of Identity-in-MotionTM systems performance requires algorithms that utilize all available information. Biometric data can be only partial or of poor quality for a variety of reasons usually dependent on the scenario in which the data was captured. For example, outdoor environments can be particularly challenging because of environmental conditions (e.g. rain, fog, smoke, dust, extraneous car lights, etc.). Retica has identity fusion algorithms which utilize not only the information from both irises but also incorporate a suite of third party face recognition algorithms. In addition we incorporate other biometric information such as interpupillary distance and periocular texture. By properly combining all available information, we have been able to produce robust identification performance under the most challenging conditions. Our futureOver the last few years Retica has worked with global partners and customers to develop a state of the art set of iris analysis tools. A comprehensive iris image library has been collected from in-house optical engines, partners and publicly available datasets. Going forward, we will continue to work with partners and customers to push the boundaries of iris identification technology in the following areas: - Speed and accuracy
- Countermeasures
- Relaxation of iris acquisition constraints
- Poor image quality mitigation
- Biometric fusion
|