The Real-Time Anomaly Detection using Computer Vision for Monitoring Key Infrastructure and Movement Patterns in Selected Areas of Niger State

Abstract

It could be said that the operational integrity and security of critical infrastructure (CI) in Niger State, Nigeria, including power installations, major roadways, and the government complexes faces significants threat from vandalism, sabotage, and security breaches. Current surveillance methods usually often fail to provide timely, actionable intelligence. This article is written to details the development and architecture of a Real-Time Anomaly Detection System (RADS) based on Computer Vision (CV) and Deep Learning (DL). The RADS is designed for a continuously, automated monitoring of selected high-value areas. It utilizes an Edge to Cloud pipeline featuring a hybrid model approach, that is YOLOv8 for point/contextual object anomaly detection (e.g., abandoned packages, unauthorized construction materials), and Convolutional Autoencoders (CAE) for collective movement pattern anomaly detection (e.g., irregular traffic flow, sudden crowd formation). By the establishing a robust baseline of 'normal' activity, the system achieves a proactive alert capability. Preliminary analysis suggests that the architecture can meet the target performance metrics of a True Positive Rate (TPR) ≥95% and a low False Alarm Rate (FAR) ≤2 per hour , significantly transforming security operations from reactive recording to intelligent, predictive threat management across Niger State.