Eye Locker

IMPORTANT: Since the hardware of today's cameras do not offer a high quality image capture for close range shots, it is recommended to use the sample images captured from special cameras with infrared light.

Download the images from:
https://mega.co.nz/#!UJUBAAjJ!VjLPPAp8TVrwTjzs_bt2mDBK2JRQRnl53BLNErqbDmI

NOTE: Within the file are images of three different subjects, with catches of left and right eye.
Left eye features ARE NOT THE SAME than the right one, although they are from the same subject!!.
(Since this is a test there is not provided with many images).

Application available in Spanish, French and German. Suitable for all screen sizes except 7 "and 10" screens.

If there is any problem with the app on your device, please report it and I will rectify immediately.

REMEMBER: THIS APP IS AN APPLICATION TEST FOR BIOMETRIC IDENTIFICATION.

If your device can take with his camera a capture that pass the quality test and afterwards identified the subject correctly, please let me know. Thank you very much.

---------------------
The possibilities offered by current technology make us more and more dependable on it and therefore we need to establish everything in connection with technological devices that we use daily. That's why the latest generation of mobile devices, called smartphones, is practically an indispensable tool for most people. These devices contain a lot of confidential information which need an effective and comfortable security system, this demand is more important day by day. From this demand has born this project, which has been based on a mobile device implemented on Android, a security system for iris recognition subject. This project has been developed to demonstrate that the implementation of high security processing systems is possible on Android devices. This can lead to future projects for the protection of sensitive data such as passwords or credit card, or lock and unlock the device.

The final results were better than expected, due to improve implementation in image processing filters and use of OpenCV libraries (better timing than a dual-core computer). For this insignificantly reduction of the computing device process and the simplification of the filters, in several tests was revealed an average of 3 seconds verification in a Samsung i9100 device (dual-core processor). These tests determined a success at 96% identification errors in images (with non-false positive), a false positive and a correct subject verification of 100% (using 150 candidates in data-base).


The project report can be found at the following link:
http://e-archivo.uc3m.es/handle/10016/16852