Enhanced fingerprint classification through modified PCA with SVD and invariant moments

Area of Science:

• Biometrics
• Computer Vision
• Machine Learning

Background:

• Accurate fingerprint identification is crucial for security.
• Existing methods face challenges with computational complexity and robustness.

Purpose of the Study:

• To introduce a novel MOMENTS-SVD vector for enhanced fingerprint identification.
• To reduce computational complexity while improving accuracy and robustness.

Main Methods:

• Feature extraction using Singular Value Decomposition (SVD) and invariant moments.
• Classification employing Euclidean distance and neural networks.
• Enhancement via modified Principal Component Analysis (PCA).

Main Results:

• The MOMENTS-SVD vector demonstrates reduced computational complexity compared to existing models.
• Comparative analysis on multiple databases (CASIA V5, FVC 2002, 2004, 2006) shows superior performance.
• Achieved enhanced accuracy and robustness against methods like ResNet and VGG19.

Conclusions:

• The MOMENTS-SVD vector offers a computationally efficient and highly accurate solution for fingerprint identification.
• This method provides a robust alternative for biometric authentication systems.
• Further validation across diverse datasets is recommended.