基于深度学习的自动扶梯乘客异常行为识别方法研究

Research on Identification Method of Escalator Passengers’Abnormal Behavior Based on Deep Learning

自动扶梯乘客异常行为识别方法的研究对保障乘客安全具有重要的意义。针对自动扶梯出入口拥堵、长时间停留等乘客异常行为缺乏有效识别和预警手段的不足,提出一种基于深度学习的自动扶梯乘客异常行为识别方法。该方法采用YOLOv4算法对自动扶梯使用场景的视频进行特征提取,识别检测区域的乘客信息;结合DeepSORT算法对检测到的乘客进行追踪和统计,构建乘客异常行为识别模型,实现乘客异常行为的识别。对4段自动扶梯监控视频的实验结果表明,该方法检测平均准确率为95.09%,能准确地识别自动扶梯出入口拥堵、长时间停留等乘客异常行为。

The research on the identification method of escalator passengers’abnormal behavior is of great significance to ensure the safety of escalators.Aiming at the shortage of effective identification and early warning methods for abnormal behaviors of escalator passengers,such as congestion and long stay at the entrance and exit of escalator,this paper proposes a method for identifying abnormal behaviors of escalator passengers based on deep learning.In this method,YOLOv4 algorithm is used to extract the features of the escalator scene video and identify the passenger information in the sensing area;Combined with DeepSORT algorithm,the detected passengers are tracked and counted,and the identification model of abnormal behavior is built to realize the identification of abnormal behavior of passengers.The experimental results of four escalator surveillance videos show that the average detection accuracy of this method is 95.09%,and it can accurately identify the abnormal behaviors of passengers such as congestion at the entrance and exit of the escalator and long stay.