CBAM-DenseNet with multi-feature quality filtering: advancing accuracy in small-sample iris recognition

Area of Science:

• Biometrics
• Computer Science
• Information Security

Background:

• Traditional password and key-based authentication are inadequate for modern information security needs.
• Iris recognition offers high security and uniqueness but current methods suffer from feature information loss.
• Single feature extraction in existing iris recognition techniques limits recognition accuracy.

Purpose of the Study:

• To propose a novel multi-feature fusion-based iris recognition method.
• To enhance the accuracy and robustness of iris recognition systems.
• To address the limitations of single feature extraction in current iris recognition approaches.

Main Methods:

• Implemented a comprehensive quality evaluation scheme for iris image filtering.
• Utilized an improved CAN network for effective image noise removal.
• Employed DenseNet for iris feature extraction, combined with a fusion space and attention mechanism (CBAM) for feature expressiveness.

Main Results:

• Validated significant improvements in recognition accuracy through experiments.
• Demonstrated enhanced robustness of the proposed iris recognition method.
• Achieved superior performance on small sample sizes and public iris databases.

Conclusions:

• The proposed multi-feature fusion method significantly improves iris recognition accuracy and robustness.
• The integration of quality evaluation, noise reduction, and advanced feature extraction enhances system performance.
• This approach offers a more secure and reliable authentication solution for the information age.