Signature Recognition and Detection of Skilled Forgeries Using Image Transformation and Multistream CNN

Abstract

Person identification through their credentials such as biometrics or signature is very important for one’s privacy and has become the most integral part for recognition. Prevention of forgeries in handwritten signatures has gain prominence in recent times. To serve this target this paper carried out Image Transformation techniques and an Artificial Intelligence model to effectively notice the differences of genuine and forged signature. Grass-fire transformations and optical flow captures the disparity in signatures. Proposed system uses Deep learning framework with ResNet 50 along with Convolutional Neural network (CNN). Comparative studies have been done using SVC2004 and SUSIG benchmark with the existing literature.