Fingerprint identification using SIFT-based minutia descriptors and improved all descriptor-pair matching

Area of Science:

• Biometrics
• Computer Vision
• Pattern Recognition

Background:

• Conventional minutiae-based fingerprint algorithms struggle with low-quality prints (e.g., damaged, small sensor overlap).
• Scale Invariant Feature Transformation (SIFT) shows promise but is unsuitable for fingerprints due to ridge patterns and computational cost.

Purpose of the Study:

• To enhance fingerprint verification accuracy and efficiency for challenging fingerprint images.
• To adapt and improve the SIFT algorithm for robust fingerprint identification.

Main Methods:

• Developed a SIFT-based Minutia Descriptor (SMD) incorporating image processing, descriptor extraction, and matching.
• Proposed an improved All Descriptor-Pair Matching (iADM) for real-time 1:N verifications.
• Integrated SMD and iADM into the Fingerprint Identification using SMD and iADM (FISiA) system.

Main Results:

• FISiA demonstrated significant accuracy improvements on representative fingerprint databases compared to conventional methods.
• The FISiA system meets real-time processing speed requirements.
• The proposed SMD and iADM effectively address SIFT's limitations in fingerprint applications.

Conclusions:

• The FISiA system offers a more accurate and efficient solution for fingerprint identification, particularly in scenarios with degraded image quality.
• The developed SMD and iADM components represent a substantial advancement in biometric security technology.
• This approach enhances the reliability and applicability of SIFT in the biometrics domain.