A Robust Deep Features Enabled Touchless 3D-Fingerprint Classification System

The exponential rise in software computing and hardware technologies has broadened the horizon for different applications in decision making to make human life efficient. Among all the major demands, security systems have always been the dominant one to ensure authenticity of data, source or certain activity. Fingerprint technology has gained wide-spread attention for personalized data, resource or activity accesses authentication. Though, numerous methods have been developed for fingerprint detection and identification, the local input environment, data suitability, distortion and hardware dependency have been the challenge to yield optimal performance. On contrary, the possibilities of touchless 3D-fingerprint identification systems have attracted scientific communities due to ease of implementation, reduced dependency on local environment and sensing hardware. In this paper deep features-based Touchless 3D-Fingerprint Classification System is proposed. In this model a transfer deep-learning model AlexNet-CNN is used for deep feature extraction and classification, which obtains 4096 dimensional deep features. The proposed approach achieves a classification accuracy of 90.20%.