Machine learning-based fingerprint pattern recognition with Mach-Zehnder interferometer

Abstract

Fingerprints serve as an excellent means of individual identification. This study employed transmission-mode image detection using a Mach-Zehnder interferometer to produce two-dimensional fingerprint images. The research aims to develop a fingerprint pattern recognition system by integrating novel algorithms for ridge orientation analysis. The fingerprint patterns considered in this study include loops, arches, and whorls. The objective is to achieve high-resolution fingerprint patterns to analyze fingerprint variability, distinguish foreground from background regions, and trace ridge line orientations. The recognition algorithm is implemented using image processing techniques, with a preprocessing workflow that includes binarization, resizing, and histogram equalization for image enhancement. After preprocessing, the image undergoes ridge orientation analysis using the gradient method, and both similarity scores and pattern categorization were successfully achieved. The algorithm's performance was evaluated through MATLAB simulations, with results demonstrating the proposed system's accuracy and reliability across various fingerprint samples. The interferometric setup was also adapted to recognize microorganisms within a microfluidic channel, highlighting its potential applications in biometrics, forensics, and medical diagnostics.