Iris recognition is a proven form of identification,
providing more accurate results than other forms of biometrics
like facial recognition or fingerprints. Yet its commercial impact
has been limited so far due to system costs and size. This report
presents the development and evaluation of an iris recognition
system that makes use of a readily available and affordable
Raspberry Pi 4. The hardware also includes near infrared image
capturing using an Arducam OV9281 NoIR camera module
and near-infrared illumination to visualize iris features that are
invisible using visible light. Iris localization is done using Python
and OpenCV and feature extraction is performed by convolving
the normalized iris image with a 2D Gabor wavelet. The resulting
iris code is matched using Hamming Distance. Results show the
system can achieve reasonable accuracy at a 0.8% false positive
rate when the iris is successfully detected but still faces challenges
with consistent iris edge detection. This holds for both images
from the CASIA v1 iris dataset and live images made by the
device. Overall, the algorithm proved to be secure and usable as
a verification system, yet more research needs to be done to be
able to use it as a means of identification.