Context-Aware Security Robot System for Dangerous Person Discrimination with Digital Twin Integration

Puntos clave

• To develop a context-aware security robot that can accurately identify dangerous individuals and visualize results in a digital twin environment.
• Integrated pose estimation and object recognition techniques.
• Developed a real-time threat detection system within a security robot.
• Utilized a Jetson Orin NX platform for on-device AI capabilities.
• Achieved a weapon detection accuracy of 95.55% based on mAP50.
• Attained a dangerous person identification accuracy of 82.5% at distances of 1-2 meters.
• Demonstrated near real-time threat recognition capabilities.

Resumen