Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (19): 274-284.DOI: 10.3778/j.issn.1002-8331.2211-0405

• Engineering and Applications • Previous Articles     Next Articles

Modeling and Recognition Method of Elevator Passenger Abnormal Behavior Based on Digital Twin

LI Conglin, WANG Qibing, LU Jiawei, ZHAO Guojun, HU Hao, XIAO Gang   

  1. 1.School of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou 310018, China
    2.School of Mechanical Engineering, Zhejiang University of Technology, Hangzhou 310014, China
  • Online:2023-10-01 Published:2023-10-01

基于数字孪生的电梯乘客异常行为建模与识别方法

李聪林,王琪冰,陆佳炜,赵国军,胡豪,肖刚   

  1. 1.中国计量大学 机电工程学院,杭州 310018
    2.浙江工业大学 机械工程学院,杭州 310014

Abstract: In order to solve the problems of abnormal cases lack and training samples scarcity in abnormal behavior recognition research, taking the abnormal behavior monitoring of vertical elevator passengers as the mission requirement, this paper puts forward the digital twin-based elevator passenger abnormal behavior modeling and recognition method.Through building the digital twin system architecture for elevator passenger behavior monitoring, the virtual and real mapping of elevator operation state and passenger behavior is completed.The abnormal behavior case of elevator passengers is built and the twin data of abnormal behavior is expanded based on the digital twin scene and the human body behavior modeling theory.Finally, the improved OpenPose is used to obtain skeletal features and the PCA-DNN is used to train the classification model to realize the abnormal passenger behavior fast recognition with fused twin data.Experimental results show that the proposed method not only reveals the convenience, efficiency and safety to create anomalous data with digital twin technology, but also verifies the feasibility, reliability and accuracy to train models with twin data.

Key words: elevator passenger abnormal behavior, digital twins, behavior modeling, data augmentation, skeleton detection, gesture recognition

摘要: 为了解决异常行为识别研究中存在的异常案例缺乏、训练样本稀缺问题,以垂直电梯的乘客异常行为监测为任务需求,提出了基于数字孪生的电梯乘客异常行为建模与识别方法。通过搭建电梯乘客行为监测数字孪生系统架构,完成了电梯运行状态与乘客行为的虚实映射。基于数字孪生场景与人体行为建模理论构建了电梯乘客异常行为案例,扩充了异常行为的孪生数据。利用改进的OpenPose获取骨骼特征,并使用PCA-DNN训练分类模型,实现了融合孪生数据的乘客异常行为快速识别。实验结果表明,该方法不仅展现了应用数字孪生技术创建异常数据的便捷性、高效性以及安全性,还验证了使用孪生数据进行模型训练的可行性、可靠性以及准确性。

关键词: 电梯乘客异常行为, 数字孪生, 行为建模, 数据增强, 骨骼检测, 姿态识别