It is highly susceptible to visual fatigue for manual surveillance of railway passenger stations, leading to decreased timeliness and accuracy. Therefore, this paper proposed a technology for operational situation awareness of railway passenger stations based on intelligent video analysis. This technology based on an artificial intelligence algorithm that deeply combined production and operation plans. It provided abnormal behaviors detecting models on platforms and analyzed passenger flow movement patterns in waiting halls respectively. It extracted wandering personnel/luggage on the platform and determined the location and trajectory of the target based on the adaptive feature pyramid two-stage cascaded network, achieving functions such as platform crossing detection, end intrusion detection, wandering in restricted areas, standardized operations, and staff attendance verification. The technology automatically generated crowd density heatmaps and movement pattern maps based on the cross-domain residual spatial fully convolutional network in situations such as the waiting hall/corridor in the foreground. Additionally, it automatically assessed queuing and inspection situations for scenarios such as nearby entrances and ticket gates. The technology intelligently adjusts the operational plans of passenger service staff, enhancing both the travel experience and waiting safety for passengers, based on the passenger flow density and abnormal situations within the station.
DUQiliang, HUANGLiguang, TIANLianfang,et al. Recognition of Passengers' Abnormal Behavior on Escalator Based on Video Monitoring[J]. Journal of South China University of Technology (Natural Science Edition),2020,48(8):10-21.
FENGTingting, GEHuayong, SUNJiahui. A Densely Connected Pedestrian Detection Algorithm Based on the Combination of SSD and FRN[J]. Information Technology and Network Security,2020,39(12):56-60,66.
LIRui, LIPing, MAXiaoning,et al. Cloud-Edge Collaboration Technology Architecture of Intelligent Analysis with Railway Videos and Images in 5G Network[J]. Railway Transport and Economy,2022,44(5):57-63.
WANGJun, XIALimin. Abnormal Behavior Detection Based on Deep-Learned Features[J]. Journal of Hunan University (Natural Sciences),2017,44(10):130-138.
[13]
LIW X, MAHADEVANV, VASCONCELOSN. Anomaly Detection and Localization in Crowded Scenes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,36(1):18-32.
[14]
SRIVASTAVAN, SALAKHUTDINOVR. Multimodal Learning with Deep Boltzmann Machines[C/OL]. (2014-01-01)[2023-05-01].
[15]
LIUP, HANS Z, MENGZ B,et al. Facial Expression Recognition via a Boosted Deep Belief Network[C/OL]. (2014-09-25)[2023-05-01].
[16]
HEK M, SUNJ. Convolutional Neural Networks at Constrained Time Cost[C/OL]. (2015-10-15)[2023-05-01].
[17]
FEICHTENHOFERC, PINZA, WILDESR P. Spatio-Temporal Multiplier Networks for Video Action Recognition[C/OL]. (2017-11-09)[2023-05-01].
[18]
HARAK, KATAOKAH, SATOHY. Learning Spatio-Temporal Features with 3D Residual Networks for Action Recognition[C/OL]. (2018-01-22)[2023-05-01].