VFRS: A video-based face recognition system to record human identity and complement the cyber identification of the infrastructure employees.
The face recognition physical sensor is complemented to audit the cyber identification of the infrastructure operators to ensure the cyber identification used to access the restricted operations is complemented with the physical identification of the personnel assigned the respective security clearance role. The detector integrates a low-cost video stream capture system that is continuously processed by the deep-learning module equipped with the database of the critical infrastructure operators. The detector notifies the central command center with the identity of the personnel gaining access to the restricted regions of the infrastructure operations. The performance of the detector is dependent upon several environmental factors including the lighting and the availability of facial features. The two components of detector correspond to the detection and the recognition. The long-range detection capability of the physical sensor extending upto 20m and has been successfully validated against indoor and outdoor lighting conditions.
The VFRS detector integrates the cutting-edge scientific components for delivering high robustness for physical identity recognition of operator personnel and external intruders. The deep learning-based component uses a customization of multi-task cascaded convolution neural networks in complementary to the Facenet component with proprietary feature extraction component. The low-complexity facial recognition component also integrates the tracking of the personnel identity to improve the robustness and accuracy of the identity recognition. The real-time computation of the multiple personnel identities complements the need for mapping cyber-physical identifications of operator personnel.
Dr. Krishna Chandramouli
Defender deliverable: D4.4, D5.4 and D5.5. All the deliverables are classified as EU RESTRICTED