Fingerprint Image Segmentation Using Machine Learning Techniques

Sophisticated Machine Learning Techniques were used to conduct an exhaustive study of fingerprint images cutoffs. An assessment and comparisons will be done with the available approaches and a new method developed that will integrate k-means grouping, deep learning and ensemble classification. The proposed approach solves common problems such as variations in quality of images and a design with a complex background while separating fingerprints and thus improves reliability and accuracy with a variety of datasets. Test results indicate that the proposed system has a better accuracy of separation and efficiency of processing compared to some best existing techniques. Identification and classification need to be very accurate in applications related to security. The study has demonstrated the potential of machine learning in fingerprint image separation and will form the basis for future research and development in biometric identification system. The proposed system provides a significant improvement in separated results and also helps advance fingerprint recognition systems in the real-life applications like forensics, security and identity verification. A very important part of fingerprint recognition is locator. The goal of fingerprint separation is the identification of the region of interest containing the specific fingerprint mark. Fingerprint image separation is very relevant to the success of automatic fingerprint identification systems. In this work, we propose a Modified Gradient Based Method to actualize area extraction. One important advantage to the approach is the ability to provide a highly refined separation rate for fingerprint images, especially under low-quality image conditions. The proposed algorithm is tested against the FVC2004 database. The results from the experiment demonstrate the improved performance of the proposed method.