Deep Face Recognition for Surveillance and Access Control

تاریخ : جون ۲۰۲۰ - جولائی ۲۰۲۱

<div><p><strong><em>For security we [humans] identify persons through their ID card but what if someone lost it or ID is compromised.</em></strong></p> <p><strong><em>This problem can be solved by Computer vision (Facial Recognition technique). A facial identification method is used for recognizing the face as well as the identity of each person by facial landmarks. People can be identified by using photos, videos, and by using real-time cameras by using facial recognition techniques. The conventional authorities might utilize versatile devices to distinguish each person when a person is a check for his identification by the Police.</em></strong></p> <p>The most important domain of the research field in biometric technology is facial recognition and its applications are widely utilized for security purposes to consumer applications, monitoring solutions to industrial efficiency, access control, and surveillance, etc.  Most of the existing facial recognition systems are based on old techniques which give poor accuracy and are vulnerable to lightning and other environmental conditions. To address these problems, we propose a <strong>Deep Face Recognition</strong> system for surveillance, access control, and other security purposes.  We also focus to make a smart attendance system in which attendance can be marked by the faces of people appearing in the camera. In this project, we prepare a custom dataset for facial recognition. We perform data augmentation to increase the variation and size of the custom dataset. We utilize a pre-trained VGG-16 model for deep face recognition. We fine-tune the VGG-16 model with our custom dataset and trained it. We implement our project for access control, surveillance, and facial attendance. In the access control setup, if the person is not recognized by our system, access to the gate will be denied.   We also maintain a database of people appearing in front of the camera during the access control setup. We record the time of entry, the name of the person, and the image of the person in our database for surveillance and record purpose. In a facial attendance setup, our system provides the ease to mark the attendance of all the registered persons appearing in front of the camera and their names and time of attendance will be stored in the database. Our system also provides the facility of registration of a new person in the existing record. We made an efficient graphical user interface to provide all the above-mentioned facilities. We also implement our complete project on an embedded platform raspberry pi 4 to shape our project in form of a portable product.</p></div>