CrisisGuard Research Paper

CrisisGuard: An Integrated AI-Based Surveillance and SOS Response System for Emergency Detection and Evacuation Guidance CrisisGuard is an AI-driven emergency response system designed to integrate fire detection, violence detection, SOS reporting, evacuation guidance, and real-time incident notifications into a unified platform. The system combines YOLO-based computer vision models with motion analysis to identify potential threats, while a sustained-threat confirmation mechanism helps reduce alerts caused by transient detections. It also supports manual and voice-triggered SOS requests, incident evidence capture, and SMS notifications to relevant responders. For evacuation guidance, CrisisGuard uses AI-assisted building floor-plan interpretation and the A* pathfinding algorithm, followed by path-smoothing techniques to generate routes toward designated exits while considering mapped obstacles and danger zones. A real-time dashboard enables incident monitoring and coordination among users, staff, dispatchers, and administrators. This paper presents the system architecture, methodology, and implementation workflow, along with an evaluation plan for assessing detection performance, alert latency, and evacuation-route validity. Quantitative performance claims are not made because controlled experimental results have not yet been established. CrisisGuard is intended as a decision-support system and does not replace human verification or established emergency procedures.