A Real-Time Smart Bus Passenger Analysis Platform for Revenue Protection and Safety Optimization in Sri Lanka

Private buses carry a large share of daily commuters in Sri Lanka, yet many operators still rely on manual ticketing, cash collection, and poorly maintained trip records. This makes it difficult to verify passenger volume, estimate expected revenue accurately, and respond promptly to unsafe driving behavior. This paper presents a real-time smart bus platform that integrates passenger monitoring, automated fare estimation, revenue forecasting, and driver supervision within a low-cost architecture designed for Sri Lankan privatebus operations. The platform uses onboard cameras to detect passengers, verify boarding and alighting events, and assign an anonymized UUID to each completed journey. These records are linked with GPS data to estimate travel distance and calculate fares automatically. The same trip records are then reused for fare analysis and short-term income forecasting, while a separate in-cabin module monitors risky driver behavior and reports alerts through an owner dashboard. Preliminary experiments using staged boarding footage, short route trials, and historical operating data produced passenger-counting accuracy in the high eighties, forecast performance suitable for trend-level planning, and dashboard updates within a near realtime range. Performance was still affected by low-light conditions, occlusion, GPS drift, and limited onboard hardware. Even so, the findings indicate that an integrated AI- and IoTbased solution can improve transparency, strengthen revenue protection, and support safer day-to-day operations without relying on smart cards or expensive proprietary infrastructure.