Automated Minutiae Extraction and Fingerprint Analysis Using Machine Learning Techniques

In the digital world, traditional methods of biometric recognition are being challenged. Traditional systems such as RFID and cards are vulnerable to fraud. However, fingerprints contain varying degrees of unique identifiers for each individual and can be exploited for precise recognition of people. Automatic feature extraction, a key part of fingerprint analysis, identifies unique ridge and edge patterns in a fingerprint image. This information can be digitally signed to enable secure identification and verification, eliminating the risk of manipulation. This work proposes a fingerprint analysis and matching system using deep learning and machine learning techniques. For feature extraction, edge detection and adaptive thresholding have been implemented. To prevent spoofing attacks, templates are generated using pattern recognition (using level 1 features fingerprints) and minutiae (also known as level 2 feature) of fingerprints. The work compares CNN, Random Forest, KNN, and SVM for fingerprint matching. The analysis of ML models reveals that CNN performs better compared to others with an accuracy of 97.6%.