摘要
目前,人脸识别技术在铁路客运车站进出站核验环节与铁路12306 APP中均得到充分应用,尤其基于人像检索的无接触出站的试点应用,显著提升了旅客出站的便利性,但旅客戴面部遮挡物出行的情况较为普遍,基于人脸识别技术的铁路车站应用受到挑战。针对基准的GaitSet步态识别算法进行改进,通过多尺度特征融合丰富步态的细节信息和语义信息的鉴别力,通过注意力机制挖掘并聚焦步态特征的关键信息,增强不同步态特征间的差异。改进的GaitSet步态识别算法,分别对开源步态数据集和自搜集的铁路场景数据进行模型训练,通过消融试验证明改进方法的有效性,其中基于铁路客运车站的试点应用,使得无接触出站能力提升2.31%,为铁路客运无接触出站研究提供参考。
Currently,face recognition has been fully implemented in entrance and exit verification processes at railway passenger stations and within the 12306 APP.In particular,the pilot implementation of face image retrieval-based contactless exit has greatly enhanced the exit convenience for passengers.Nonetheless,it is common for passengers to wear face coverings,which presents challenges for face recognition-based applications in railway stations.The baseline GaitSet algorithm for gait recognition was improved.The improved algorithm incorporates multi-scale feature fusion to enhance the discriminative power of detailed and semantic information on gait.Additionally,an attention mechanism was employed to extract and focus on the key information on gait features,thereby improving inter-feature discriminability.The improved GaitSet algorithm was trained on both an open-source gait dataset and a self-collected dataset from real railway scenarios.An ablation experiment was performed to validate the effectiveness of the improved method.Additionally,a pilot implementation in railway stations demonstrated a 2.31% improvement in contactless exit capability,providing references for future research on contactless exits at railway passenger stations.
作者
李贝贝
阎志远
戴琳琳
刘相坤
车儒平
LI Beibei;YAN Zhiyuan;DAI Linlin;LIU Xiangkun;CHE Ruping(Institute of Computing Technology,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;Railway Travel Technology Corporation Limited,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China)
出处
《铁道运输与经济》
北大核心
2025年第7期150-158,共9页
Railway Transport and Economy
基金
中国铁道科学研究院集团有限公司科研项目(2023YJ138)。
