TOUCHLESS FINGERPRINT IDENTIFICATION USING DEEP LEARNING

Abstract

Fingerprint identification technology is used as one of the most reliable technologies for authenticating people. The main reason for its popularity is the high reliability and unique ness. Fingerprint identification technology currently on the market is based on contact-based fingerprint scanners or 2D enrollment fingerprint sensors. In the early stages of research, many basic feature extraction techniques were used, such as dividing samples into discrete marker skeletons or tediously calculating the shape of adjacent grooves. Obtaining fingerprint image is one of the major stumbling blocks. This is a problem faced by all the researchers, and further studies and techniques are being developed in this area. The contactless fingerprint recognition system is a new choice for the old conventional touch-based fingerprint recogni tion system. In this work, IIT-Bombay Touchless and Touch Based Fingerprint Database is used for classification which contains 200 subjects. Deep learning is currently popular in computer vision and pattern recognition. CNNs and deep learning models have proven to be extremely effective in solving image processing problems and provide breakthrough results. In this work, pre-trained deep learning models such as VGG-16, VGG-19, Inception-V3 and ResNet-50 architecture by using transfer learning. The results obtained where promising in which VGG-16 architecture performed better among other pretrained model architecture with an accuracy of 98% for training the preprocessed datasets and 93% for unprocessed datasets.