Real Time Anomaly Detection with Video Storage Optimization for Smart CCTV Surveillance

Traditional surveillance cameras had limited recording and storage capacity, restricting their ability to retain recorded videos for a longer duration. Advancements in Closed Circuit Television (CCTV) technology and the integration of the Internet of Things (IoT) have significantly improved surveillance performance efficiency. This work demonstrates a methodology for secure surveillance by implementing Convolutional Neural Network (CNN) based real-time anomaly detection and optimization of the recorded footage simultaneously. The proposed system also offers automated alert notifications to the user for anomalies detected, like theft, unlawful entrance, hostile behavior, and other suspicious actions. This system is capable of handling large volumes of surveillance data effectively without compromising the accuracy of anomaly detection. Experimental results based on the F1-confidence ratio and precision-recall curve validate the effectiveness of detecting weapons and suspicious behavior in real-time surveillance scenarios.