Cross-Sensor Fingerprint Matching Using Siamese Network and Adversarial Learning.

Area of Science:

• Biometrics and Pattern Recognition
• Computer Vision and Machine Learning

Background:

• Fingerprint verification is crucial for identity authentication worldwide.
• Cross-sensor matching presents significant challenges due to varying sensor technologies and enrollment/query phase discrepancies.
• Existing systems struggle with interoperability, necessitating robust automatic solutions.

Purpose of the Study:

• To propose an efficient, robust, and automatic system for cross-sensor fingerprint matching.
• To address the sensor interoperability challenge in biometric authentication.

Main Methods:

• A novel cross-matching system, SiameseFinger, is developed utilizing a Siamese network architecture.
• Feature extraction is performed using the Gabor-HoG descriptor.
• The Siamese network is trained through adversarial learning techniques.

Main Results:

• SiameseFinger was evaluated on two benchmark public datasets: FingerPass and MOLF.
• Experimental results demonstrate that SiameseFinger achieves performance comparable to current state-of-the-art methods.
• The system shows promise in overcoming cross-sensor matching limitations.

Conclusions:

• The proposed SiameseFinger system offers a viable solution for cross-sensor fingerprint verification.
• Adversarial learning within a Siamese network framework effectively handles sensor interoperability.
• This research contributes to more robust and versatile biometric authentication systems.