Using Motion Detection and Facial Recognition to Secure Places of High Security: A Case Study at Banking Vaults of Ghana

Abstract

Motion Detection and Facial recognition (MD&FR) have been extensively studied to improve surveillance services in areas where motion detections and human identification are needed. As crime techniques keep improving, surveillance technologies must also advance in the same manner, and MD&FR is the reliable surveillance technology in current literature. Despite the conceivable potentials of integrating MD&FR technologies into current surveillance systems (SS), the available systems have deployed MD-based and FR-based SS in isolation. In this paper, we proposed a Smart Surveillance Systems (SSS) functional framework, the design Flowchart of this SSS and a software approach to implement the SSS using MD&FR techniques in OpenFace and OpenCV Facial Recognition (FR) and Motion Detection (MD) libraries, respectively. Extreme Learning Machine (ELM) was the Facial Recognition (FR) algorithm used due to its robustness and higher accuracies under variable regularization factors (RF). Contrasting with the current CCTV Camera-based surveillance systems used in Ghanaian banking Vaults, this system augmented security when it was deployed. Thus, our system saves much storage space because it records automatically only during motion detections, logs all detected faces and alert instantly if intrusion is detected. The SSS worked best with CCTV Cameras and also operated efficiently with lower pixel-rated (cheap) cameras, such as WebCams and Infinix Note3 under variable RF. This system is recommended for policy consideration in banks and other firms to minimise security systems’ overheads and boost efficiency of SS.