An Architecture for Human Action Recognition in Smart Cities Video Surveillance Systems

Abstract

Facial recognition systems are part of our daily lives. The face is already used to unlock mobile phones, withdraw cash at ATMs, pay at establishments, perform checks at airports or identify suspects at large events such as football matches or concerts. In Smart Cities there is a large amount of information about the environment, so it is very interesting to apply techniques to characterize the different domains and the detection of events and certain situations. In this work we propose a system for access control through facial recognition and data extraction of the individual through an identification card and deep learning. The architecture will be easily scalable and adaptable to new situations and fields of application. Therefore, our system will identify human features in video surveillance sequences using deep learning techniques to support video monitoring systems. The results obtained with different databases provide accuracies over 90%, which proves the validity of our proposal.