The Classification of Arch Fingerprint Using Mathematical Model and Deep Learning Features Selection

Fingerprints are unique patterns, made by friction ridges and furrows, which appear on the pads of the fingers and thumbs. The analysis of fingerprint is valuable tool to identify suspects and crimes. There are three basic types of fingerprints, one of them is the arch fingerprint. In this paper, the classes of the arch fingerprint are modelled in a dynamical system. The global dynamics and the existence and stability of equilibria are studied. The orientation image of the arch fingerprint is visualized in smooth deformation of the phase portrait of a planar system. Most fingerprint datasets are not categorized to be retained by deep learning computer tools that allow to classify a new fingerprint input image to its class, so finding a dynamical system to categorize fingerprint image data set allows deep learning computer science program to be retrained with more accuracy. Convolutional Neural Networks (CNNs) architectures are computerized machine learning tool that used a deep learning for classifying images. CNNs are eligible of automatically extracting and learning features from any categorized dataset. VGG16, GoogleNet, Resnet, and Alexnet are CNN architecture that are retrained over NIST Special Database (SD) 302d fingerprint dataset.