Design of a Multi-source Perception Onboard Safety Management System for Rail Transit Production Vehicles

Heavy-haul railway production vehicles are diverse in type, widely distributed, and operate with irregular schedules, making it difficult to effectively monitor driver behavior. To address this issue, an integrated onboard safety management solution combining facial recognition, multi-feature fatigue analysis, high-precision positioning, and reliable 4G transmission is proposed. The system consists of three components: an onboard terminal, a central management platform, and a mobile app, integrating driver identity verification, mid-journey driver change alerts, self-cleaning alcohol lockout, dual-camera fatigue detection, and electronic fence alarms into a closed-loop end-to-cloud control framework. A three-month field test was conducted on two off-road vehicles and one minibus operated by Baoshen Railway, involving over 500 startup verifications and more than 400 hours of driving monitoring. Test results show that facial recognition accuracy reaches 98.5%, alcohol detection error does not exceed 8%, fatigue detection accuracy exceeds 92%, alarm response time is ≤1.2 seconds, and system availability is 99.996%, all meeting design requirements. This solution effectively prevents unauthorized driving, drunk driving, and fatigued driving, providing a scalable technical approach for intelligent safety management of railway production vehicles.