Turnstile Status and Passenger Posture Based Passenger ticket Evading Action Surveillance System

Abstract Ticket evasion has deteriorated with the development of automation in smart metro stations. To reduce management costs and improve To reduce management costs and improve operational efficiency, the paper designs and implements the Passenger ticket evading Action Surveillance System (PASS). The system utilizes YOLOv5 network to identify the wing-type turnstile and generates surveillance areas accordingly. The system designs a tracking The system designs a tracking algorithm to build the passenger posture and displacement sequences using the Openpose network. The system designs a tracking algorithm to build the passenger posture and displacement sequences using the Openpose network. According to the test result, PASS improves by 6.0-41.2% in detection accuracy and reduces by 6.1–58.3% in false positive rate compared to traditional purely pose estimation. According to the test result, PASS improves by 6.0-41.2% in detection accuracy and reduces by 6.1–58.3% in false positive rate.