Abstract:The growing concern over high-profile violence has led to a widespread adoption of intelligent smart video surveillance systems in educational institutions. However, accurately identifying abnormal events in the video stream remains a complex task due to the absence of a definitive generic definition for abnormal actions, heavily reliant on contextual factors. In this paper, we propose a fog-assisted abnormal motion recognition system utilizing an ontology-based semantic algorithm to bolster security within university campuses under real-time surveillance. Our ontology-based semantic analysis is meticulously designed to characterize intricate spatio-temporal interactions and contextual relationships in the video scene, thereby significantly improving the sensitivity in distinguishing abnormal events based on their severity level. To ensure swift responses to specific abnormal events, the video streams captured through surveillance cameras are processed at the edge of the network. Prior to semantic analysis, two crucial steps—foreground object segmentation and object tracking—are executed to streamline the detection process. These steps involve segmenting the target foreground using a connected components labeling algorithm and tracking motion patterns based on velocity, direction, and acceleration, employing the Kalman filter. The rule-based reasoning of the ontology accurately defines abnormal conditions in the video scene, providing a clearer understanding of the decision-making process. Furthermore, we incorporate a context-aware refinement step aimed at enhancing detection accuracy. This step distinguishes abnormal events based on their severity and generates corresponding alerts. We conducted extensive experiments and evaluated the proposed system using various measures, demonstrating its potential in real-time video analysis. The results showcase an impressive prediction success ratio of 0.989, affirming the reliability and robustness of our system in detecting abnormal events.

Fog-Assisted Abnormal Motion Detection System: A Semantic Ontology Approach

Abnormal Behavior Detection Based on Regional Optical Flow

Automatic Detection of Anomalies in Video Surveillance using Artificial Intelligence

Abnormal Event Recognition: A Hybrid Approach Using SemanticWeb Technologies

Unveiling Anomaly : Empowering Video Surveillance through Intelligent Anomaly Detection

A novel approach for enhanced abnormal action recognition via coarse and precise detection stage

A high semantic representation for abnormal events detection in crowded scenes

Motion anomaly detection and trajectory analysis in visual surveillance

Autonomous Anomaly Detection System for Crime Monitoring and Alert Generation

A novel framework and concept-based semantic search Interface for abnormal crowd behaviour analysis in surveillance videos

Object-Oriented Anomaly Detection In Surveillance Videos

Ontology-based Surveillance Video Archive and Retrieval System

Deep Learning Approach for Suspicious Activity Detection from Surveillance Video

Abnormal Activity Recognition in Private Places Using Deep Learning: A Survey

Abnormal events detection using spatio-temporal saliency descriptor and fuzzy representation analysis

Video anomaly detection and localization based on appearance and motion models

Pose-oriented scene-adaptive matching for abnormal event detection

An Advanced Motion Detection Algorithm With Video Quality Analysis for Video Surveillance Systems

Detecting Abnormal Activities In Video Based Fuzzy Method

A real time crime scene intelligent video surveillance systems in violence detection framework using deep learning techniques

Real-Time Anomaly Detection in Video Streams