Estimating Passenger Number in Trains Using Multi-Source Data

Identifying passenger numbers in trains is significant for the regular operation and emergency disposal of urban rail transit. Using manual and weighing systems to capture the boarding and alighting passenger number is expensive and inaccurate. Manual surveys usually count boarding and alighting passengers propagate along train runs, so the counting errors are typically flow-dependent. The weighing data captured from onboard units are unstable and unreliable. This paper proposes a data-driven fusion method for passenger numbers in train estimation in public transport using HD camera data, automatic fare collection data, and train timetable data. It makes the best use of the complementary advantages of these two data sources in inferring the passenger travel paths within the metro system and the passenger streamline in transfer stations (camera data). We propose a train passenger load estimation model to estimate the passenger number in trains by theoretically deriving every passenger’s itinerary.