AI-Assisted Anomaly Detection for Security and Crowd Environments: A Comparative Study of Deep Learning and Edge-Enabled Surveillance Systems

作者：Seema J K, Akash Pradeepan, Aryan K S, Akileshwaran S, Nithin Bharadwaj B P · 发表于：Zenodo (CERN European Organization for Nuclear Research) · 年份：2026 · DOI：10.5281/zenodo.22065940 · 研究领域：Anomaly Detection Techniques and Applications、Network Security and Intrusion Detection、Advanced Data and IoT Technologies

The swift expansion of smart surveillance systems has created unparalleled demand for automated technologies that can identify unusual behavior, maintain security, and handle crowd dynamics in real-time. Conventional rule based surveillance methods are becoming less effective in intricate, ever-changing settings like airports, public transport centers, and large event spaces, where the overwhelming data volume and unpredictability of human actions render manual oversight impractical and prone to mistakes. The combination of progress in deep learning, edge computing, and sensor miniaturization now allows for the implementation of smart systems capable of simultaneously processing high- resolution video streams, sensor data, and contextual metadata, creating new opportunities for proactive threat prevention and public safety management. This document outlines a thorough AI-supported framework that combines deep learning, computer vision, and multimodal sensor integration to tackle anomaly detection, security implementation, and crowd environment evaluation within a cohesive structure. The suggested system is intended to function as a modular, scalable platform that can be utilized across various hardware setups without the need for specialized high performance computing resources. By consolidating three formerly isolated tasks—anomaly detection, perimeter security, and crowd flow estimation— into one inference pipeline, the framework greatly decreases operational overhead and e...