A Robust Approach on Rectification of Fingerprint Distortions Using MLM-SVM Classifier

Abstract

In the current Internet world engagement of trustable users is mandatory. Various authentication techniques are adopted in which fingerprint authentication is frequently used. Due to various reasons elastic distortion of finger print become a serious issue during identification of systems. The distortion in the fingerprints enable the system to make negative recognition in many applications. Due to duplication negative authentication create serious issues. The application of negative recognition provides additional weightage to the authentication process. The proposed system is focused on creating a robust methodology to correct the skin deformation. The presented approach detects the finger print orientation and maps the feature vectors using Hough transform and Multi-Level Morphology (MLM) for detection of distortion level. SVM classifier is used to train the reference images of fingerprints to manage the distortion pattern. During the testing phase the real time fingerprint images are fetched to check the level of distortion. The proposed algorithm creates distortion map where the real time images classified with the SVM classifier. In terms of improving the privacy preserving models where the fingerprint is utilized for authentication the proposed approach performs better comparing with the state of art approaches.