Toward real-time congestion measurement of passenger flow on platform screen doors based on surveillance videos analysis

Most metro stations lack automatic methods for monitoring the congestion and order of alighting and boarding passengers at each door on the platforms. Existing methods attempt to analyze historical data and predict the potential congestion but perform poorly on the burst of passengers. Furthermore, the number of passengers is the most direct indicator, but it is rough and ignores the interaction between passengers. To precisely measure the congestion of passenger flow, this paper proposes a real-time congestion measurement of passenger flow based on surveillance video analysis. First, a passengers trajectory extracting module is proposed to extract the passengers' moving tracks from the surveillance videos. Then, the passenger flow of a train gate is treated as a bidirectional flow that goes through a bottleneck. The congestion is measured from three aspects and calculated in entropy: (1) position entropy: considering the effect of counterflow passengers (2) speed entropy: considering the effect of speed changes between passengers, and (3) angle entropy: considering the effect of the direction of passenger movement. Moreover, to improve the sensitivity of the entropy to the crowding behavior of bidirectional flows, the K-Means clustering method is applied to the speed entropy calculation process. Finally, an entropy for measuring the congestion and order of alighting and boarding passengers can be obtained. The proposed entropy measurement method is applied to the real-world data of a Guangzhou Metro station and effectively reflects the actual congestion situation. (c) 2023 Elsevier B.V. All rights reserved.