Real-Time Identification of Persons by Iris Recognition

Recognition of persons by analysis of the iris texture is accepted as one of the most efficient biometric methods for identification, with the objective of controlling access of individuals to buildings, offices, equipment and other protected resources.

The usual control access methods involve memorizing passwords or alphanumeric codes, which can be easily forgotten or, in the worst case, stolen. This is why biometric systems based on morphologic features of persons are being increasingly considerated as a solution for different applications.

In particular, recognition of persons by iris has the advantages of being non-invasive, not requiring physical contact with any device of any kind and having enormous reliability.

The pattern of the iris is unique for every individual, it is highly differentiable between individuals (low amount of false positives) and highly repeatable (low amount of false negatives), unlike other biometric features such as the face and even fingerprints.

Implemented system

In this work, we've built a complete and automatic system of identification of persons based on iris recognitions. We developed optimized algorithms for working in real-time environments.

The system is composed of an infrared video camera which is used for capturing images of the eyes of the individuals, and the software needed to process such images. The system works in real time, which minimum interaction between the operator and the system.

The system has been tested against a database of over 1000 images of eyes, and it achieved a correct recognition rate of over 99%.

How the system works

The system works by capturing a stream of video and processing it in real-time. When a user puts his eye close to the video camera, the system detects this and tries to capture a sharp image of the eye of the user.

The image of the eye is then analysed using image processing algorithms. The texture of the iris is extracted and an unique code is generated from that texture. The code is then compared against codes previously stored in a database, and when two codes match, the person is identified.

The following video illustrates how the system works: