Fingerprint Reconstruction: Convolutional Neural Network Based Approach to Improve Fingerprint Recognition
Milind B. Bhilavade, K. S. Shivaprakasha, Meenakshi Ravindra Patil, Lalita Sunil Admuthe, Aditya N. Magdum · 2024
Convolutional Neural Networks (CNNs) in fingerprint reconstruction is a crucial aspect of biometric authentication systems. Traditional methods, such as minutiae extraction and matching techniques, struggle with noisy and distorted fingerprint images. The advent of CNN s offers a promising avenue for improving the accuracy and robustness of fingerprint recognition systems. CNNs, which can automatically learn hierarchical features, have demonstrated remarkable success in various image recognition tasks, making them an attractive choice for fingerprint recognition. This paper discusses the application of CNN s in fingerprint recognition, exploring the methodology, results and analysis of the proposed model. The contributions of this research include building a CNN model for fingerprint reconstruction, evaluating the performance of the proposed system compared to traditional methods, and analyzing the robustness of the CNN-based system under varying conditions. Deep learning methods are proved ideal for reconstructing damaged fingerprint images due to poor skin condition, large cuts, wet fingers, or poor scanning, with training accuracy of 98.75 % and validation accuracy of 99.02%.