Loading...
Please wait, while we are loading the content...
Similar Documents
Emerging challenges in iris recognition
| Content Provider | Indraprastha Institute of Information Technology, Delhi |
|---|---|
| Author | Arora, Sunpreet Singh |
| Abstract | Iris recognition has been used to recognize cooperative subjects in controlled environments. With low cost iris scanners, the technology will witness broader applications and may be confronted with newer challenges. In this research, we have proposed and investigated multiple challenges, namely matching iris images captured before and after alcohol consumption, iris camera inter- operability, matching iris images acquired before and after administration of pupil dilation eye drops and matching pre and post cataract surgery irides. Due to alcohol consumption, the pupil dilates/constricts which causes deformation in iris pat- terns, affecting iris recognition performance. Variations in iris images captured using different iris cameras also affect the process of iris recognition. Similarly, pupil dilation using eye drops as well as interventions during cataract surgical procedures affect the iris patterns. The experi- ments performed on the databases collected as a part of this research show that in matching pre and post alcohol consumption images, the overlap between genuine and impostor match score distributions increases by approximately 20%. This overlap increases by 60% when matching pre and post cataract surgery images, and by 75% when the gallery consists of images captured before pupil dilation while the probe is an image acquired after pupil dilation using eye drops. It increases by 18% when matching images of the same iris acquired using two different iris cameras. The recognition accuracies of commercial iris recognition systems also reduce signifi- cantly for such cases. These results suggest that identification using iris biometrics is adversely affected due to these emerging covariates. This research also proposes extent of change algo- rithm to measure useful area of iris images that can be helpful if pupil dilation and constriction cases. Further, an active learning based iris camera classification framework is proposed to address iris camera interoperability. This camera classification is used to perform selective iris image enhancement. Experiments show a significant improvement in cross sensor iris recognition accuracy (obtained using commercial systems) using the proposed approach. |
| File Format | |
| Language | English |
| Access Restriction | Authorized |
| Subject Keyword | Biometrics Iris Recognition Image Analysis Machine Learning |
| Content Type | Text |
| Educational Degree | Bachelor of Technology (B.Tech.) |
| Resource Type | Thesis |
| Subject | Data processing & computer science |
