Analysis of Fingerprint Image Recognition using Deep Residual Convolutional Neural Network

One biometric tool that offers better accuracy and lower cost than other biometric modalities is fingerprint authentication. However, as fingerprints become more prevalent than other biometrics and the amount of data increases, the problem of fingerprint image identification can only be solved by high processing power. Researchers delved into various techniques to enhance the accuracy and reliability of fingerprint analysis. We call these techniques Content-based Image retrieval. New techniques called Deep Learning are proposed and no longer need image analysis. In this study, we have analyzed fingerprint matching with various Residual Networks to obtain high accuracy in a short time. ResNet is considered an ideal architecture for fingerprint identification. The FVC2000 dataset was used as the test data, which was trained on different CNN architectures, namely ResNet-34 and ResNet-50, to enable faster, more accurate, and concise processing of the findings. Experimental results show that ResNet-34 takes a long time to be trained but reaches a perfect accuracy of 100%, and the fastest matching time is 0.042 seconds.