Year: 2024
Type: article
Abstract: As security challenges evolve, traditional surveillance systems often fall short in effectively identifying and responding to real-time threats. This paper introduces SurveilGuard, a novel, AI-powered surveillance framework designed to autonomously detect, report, and respond to abnormal activities such as fighting, smoking, hugging, and other predefined behaviors in real time. Leveraging motion-triggered camera activation via the ESP32-CAM and powerful anomaly detection models like YOLO for activity recognition, SurveilGuard offers a seamless integration of video capture, behavior analysis, and incident reporting. When abnormal activities are detected, the system sends real-time alerts using Twilio and automatically generates detailed reports enriched with video evidence and contextual data using CLIP for image-text matching. These reports are securely stored and easily accessible via a web interface for authorized personnel, enhancing situational awareness and operational response. Additionally, Power BI is employed for data visualization, allowing for comprehensive reporting and interactive dashboards. SurveilGuard represents a significant advancement in automated surveillance, offering a scalable solution for real-time security monitoring with minimal false alarms, empowering security teams to respond quickly and effectively.
Authors: S. Hariprasath, A. Harihara Sudhan, S. Keerthi Vasagan, P. Roy Sudha Reetha
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4405615682