A Fingerprint Recognition System for Noisy Or Partial Fingerprint by Approximate Auto-Fill Method

Fingerprint recognition systems have emerged as a popular and reliable biometric authentication method, serving a wide range of applications, from securing sensitive information to identifying criminals. However, traditional methods face challenges when dealing with poor-quality fingerprint images containing missing portions. Matching such images accurately with complete fingerprints becomes a daunting task. In this paper, we propose a novel preprocessing technique that automatically fills small missing regions in fingerprint images without distorting their inherent features. This preprocessing approach significantly improves the overall accuracy of fingerprint recognition systems. Furthermore, we employ a Convolutional Neural Network (CNN) model for classification tasks in fingerprint recognition. The proposed methodology is validated using the Socoto Coventry Dataset, where the preprocessing technique is implemented and evaluated. Our experimental results demonstrate that the proposed preprocessing technique achieves remarkable accuracy, with a recognition rate of 94.76%. By effectively addressing the challenge of missing regions, our approach enhances the reliability and robustness of fingerprint recognition systems. The combination of auto-fill preprocessing and CNN classification offers a potent solution for enhancing biometric authentication in various security-sensitive domains. This research contributes to the advancement of fingerprint recognition systems, providing improved accuracy and performance in real-world applications.