基于深度学习的地铁站点环境满意度评估算法  

Deep learning-based algorithm for environmental satisfaction assessment at metro stations

作　　者：王佳丽 杨鹏 宋程程 王欣 毕慧博 陈艳艳[1] 章明晖 WANG Jiali;YANG Peng;SONG Chengcheng;WANGXin;BI Huibo;CHEN Yanyan;ZHANG Minghui(Beijing Municipal Key Laboratory of Traffic Engineering,College of Metropolitan Transportation,Beijing University of Technology,Beijing 100124,China;China Waterborne Transport Research Institute,Beijing 100088,China;Quantutong Location Network Co.,Ltd.,Beijing 100124,China;Beijing Metro Operation Co.,Ltd.Communication Signal Branch,Beijing 100082,China)

机构地区：[1]北京工业大学北京市交通工程重点实验室,北京100124 [2]交通运输部水运科学研究院,北京100088 [3]全图通位置网络有限公司,北京100124 [4]北京市地铁运营有限公司通信信号分公司,北京100082

出　　处：《现代城市轨道交通》2024年第10期107-116,共10页Modern Urban Transit

基　　金：国家重点研发计划课题(2020YFB1600703)。

摘　　要：为全面评估北京地铁乘客对站内环境的满意度,对地铁站内环境图像进行经过隐私保护处理的数据采集,通过问卷调查和随机森林算法,确定影响地铁站内环境满意度的关键指标及其权重,利用被试人员的评估数据,建立包含满意度评分的环境满意度评估数据集。通过学习环境图像与满意度评分的关系,构建深度卷积神经网络模型。为增强模型泛化性,采用迁移学习方法。在北京地铁14号线测试集上,模型评估准确率达到96%,在未经预训练的地铁7号线上迁移学习模型达到94%的准确率。实验结果表明,该模型可准确反映乘客对地铁环境在可达性、安全性、舒适性和愉悦性方面的体验,为地铁服务的优化提供有力参考。To comprehensively assess the passenger satisfaction of the environment in Beijing metro stations,firstly,the data collection of the environment images in metro stations with privacy-protected processing is carried out,and through questionnaire surveys and random forest algorithms,the key indexes and their weights affecting the satisfaction of the environment in metro stations are determined,and the assessment data of the subjects are utilized to establish the environmental satisfaction assessment dataset which contains the satisfaction scores.Next,a deep convolutional neural network model is constructed by learning the relationship between environmental images and satisfaction scores.To enhance the model generalization,migration learning method is used.The model assessment accuracy reaches 96%on the test set of metro line 14,and the migration learning model achieves 94%accuracy on the metro line 7 which is not pretrained.The experimental results show that the model can accurately reflect passengers'experience of the metro environment in terms of accessibility,safety,comfort and pleasantness,providing a powerful reference for the optimization of metro services.

关 键 词：地铁 深度学习 环境满意度评估 图像分类 

分 类 号：U491.1[交通运输工程—交通运输规划与管理] X822[交通运输工程—道路与铁道工程]