Abstracted and indexed in:
Future:
An iris recognition is a biometric way of identifying people in the ring-shaped portion of the eyeball surrounding the pupil. An iris recognition is used in biometrics because each iris is unique to an individual. Unfortunately, even though researchers have considered various approaches to improve the detection of iris recognition, obtaining higher accuracy remains a challenging task. More specifically, the major drawbacks contributed by the poor quality of images such as blur, lighting infection, and data scarcity. Therefore, in this work, we proposed the utilization of semantic segmentation and data augmentation approach to enhance the iris detection capability in terms of accuracy. The semantic segmentation (SS), a part of Mask R-CNN, is applied to overcome the image quality limitation. This approach partitions an image into multiple image segments known as image regions to differentiate dissimilar objects in an image using pixel level. Subsequently, using the data augmentation (DA) approach, new data is derived artificially from existing data that has been effective in improving the model generalization and precisely solving issues of data scarcity. The proposed model namely SS+DA has been evaluated using benchmark datasets known as CASIA and IITD. The experiment result shows that the proposed method is able to obtain an above 99% accuracy rate for both the CASIA and IITD datasets.
Warusia Yassin
Universiti Teknikal Malaysia Melaka, Melaka, Malaysia
Mohd Faizal Abdollah
Universiti Teknikal Malaysia Melaka, Melaka, Malaysia
Sasikumar Gurumoorthy
J.J. College of Engineering and Technology, Trichy, India
Kumar Raja
REVA University, Bengaluru, India
Izzatul Nizar
Universiti Teknikal Malaysia Melaka, Melaka, Malaysia
POWERED BY
PARTNERS
