Passenger attribute recognition method for railway passenger stations based on AL-Transformer model
-
摘要: 随着铁路运力的不断提升，旅客在铁路客运站内候车的频次和时间也在不断增加，为主动挖掘候车旅客的个性化需求，提出一种基于AL-Transformer（Attribute Localization—Transformer）模型的铁路客运站旅客属性识别方法。AL-Transformer模型基于Swin Transformer主干网络提取进站旅客的结构化信息，通过掩码对比学习（MCL ，Mask Contrast Learning）框架抑制特征区域相关性，获取到更有辨识度的属性区域；通过属性空间记忆（ASM ，Attribute Spatial Memory）模块选取更加可靠、稳定的属性相关区域。在中国铁路兰州局集团有限公司白银南站试用的效果表明，该方法可有效识别旅客属性，为客运站工作人员推送更有针对性的信息，提升客运站的旅客服务质量，保障旅客候车安全。
-
关键词:
- 属性识别 /
- AL-Transformer模型 /
- 掩码对比学习（MCL） /
- 属性空间记忆（ASM） /
- 旅客异常行为
Abstract: With the continuous improvement of railway transportation capacity, the frequency and time of passengers waiting for trains in railway passenger stations are also increasing. To actively explore the personalized needs of waiting passengers, this paper proposed a passenger attribute recognition method for railway passenger station based on the AL-Transformer (Attribute Localization Transformer) model. The paper used AL-Transformer model based on the Swin Transformer backbone network to extract structured information of passengers entering the stations, suppressed feature region correlation through the Mask Contrast Learning (MCL) framework to obtain more recognizable attribute regions, and used Attribute Spatial Memory (ASM) module to selecte more reliable and stable attribute related regions. The trial results at Baiyin South Station of CHINA RAILWAY Lanzhou Group show that this method can effectively identify passenger attributes, push more targeted information for station staff, improve the quality of passenger service at the station, and ensure the safety of passenger waiting. -
-
表 1 多种方法的性能比较
方法 PA100k PETA mA Accu Pre Recall F1 mA Accu Pre Recall F1 位置信息嵌入 80.68 77.08 84.21 88.84 86.46 86.30 79.52 85.65 88.09 86.85 视觉属性聚合 - - - - - 84.59 78.56 86.79 86.12 86.46 视觉注意一致 79.04 78.95 88.41 86.07 86.83 83.63 78.94 87.63 85.45 86.23 本文方法 84.61 78.86 84.11 91.03 87.43 89.54 80.75 86.15 90.04 88.05 
下载: 导出CSV
表 2 在PETA和PA100K上的消融实验
方法 PA100k PETA mA Accu Pre Recall F1 mA Accu Pre Recall F1 Swing Transformer网络 82.82 81.47 89.08 88.88 88.98 87.20 80.17 86.54 88.73 87.62 Swing Transformer主干网络+MCL框架 83.21 81.70 89.18 88.99 89.09 87.67 71.65 76.81 89.01 82.46 Swing Transformer主干网络+ASM模块 84.00 77.00 82.85 90.08 86.32 89.26 79.55 84.83 89.92 87.30 本文方法 84.61 78.86 84.11 91.03 87.43 89.54 80.75 86.15 90.04 88.05
下载: 导出CSV
-
[1] 贾 健,陈晓棠,黄凯奇. 监控场景中的行人属性识别研究综述[J]. 计算机学报,2022,45(8):1765-1793. DOI: 10.11897/SP.J.1016.2022.01765 [2] Kumar N, Berg A, Belhumeur P N, et al. Describable visual attributes for face verification and image search[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(10): 1962-1977. DOI: 10.1109/TPAMI.2011.48
[3] Kumar N, Belhumeur P, Nayar S. FaceTracer: a search engine for large collections of images with faces[C]//10th European Conference on Computer Vision, 12-18 October, 2008, Marseille, France. Berlin: Springer, 2008: 340-353.
[4] Bourdev L, Maji S, Malik J. Describing people: a poselet-based approach to attribute classification[C]//2011 International Conference on Computer Vision, 06-13 November 2011, Barcelona, Spain. New York: IEEE, 2011: 1543-1550.
[5] Liu X H, Zhao H Y, Tian M Q, et al. HydraPlus-Net: attentive deep features for pedestrian analysis[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision, 22-29 October 2017, Venice, Italy. New York: IEEE, 2017: 350-359.
[6] Liu P Z, Liu X H, Yan J J, et al. Localization guided learning for pedestrian attribute recognition[C]//29th British Machine Vision Conference, 3-6 September 2018, Newcastle, UK. BMVA Press, 2018.
[7] Sarafianos N, Xu X, Kakadiaris I A. Deep imbalanced attribute classification using visual attention aggregation[C]//Proceedings of the 15th European Conference on Computer Vision, 8-14 September, 2018, Munich, Germany. Berlin: Springer, 2018: 708-725.
[8] Tang C F, Sheng L, Zhang Z X, et al. Improving pedestrian attribute recognition with weakly-supervised multi-scale attribute-specific localization[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision, 27 October 2019-02 November 2019, Seoul, Korea (South). New York: IEEE, 2019: 4996-5005.
[9] Liu Z, Lin Y T, Cao Y, et al. Swin transformer: Hierarchical vision transformer using shifted windows[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 10-17 October 2021, Montreal, Canada. New York: IEEE, 2021: 9992-10002.
[10] Deng Y B, Luo P, Loy C C, et al. Pedestrian attribute recognition at far distance[C]//Proceedings of the 22nd ACM International Conference on Multimedia, November 2014, 0rlando, USA. New York: ACM, 2014: 789-792.
[11] Li D W, Zhang Z, Chen X T, et al. A richly annotated pedestrian dataset for person retrieval in real surveillance scenarios[J]. IEEE Transactions on Image Processing, 2019, 28(4): 1575-1590. DOI: 10.1109/TIP.2018.2878349
[12] Guo H, Zheng K, Fan X C, et al. Visual attention consistency under image transforms for multi-label image classification[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15-20 June 2019, Long Beach, CA, USA. New York: IEEE, 2019: 729-739.
-
期刊类型引用(6)
1. 苗守庭，景晓志，冉泽宇，张志强，于茜，李红艳. 数字人民币智能合约在铁路客票+业务支付结算的应用研究. 铁道运输与经济. 2024(03): 125-133 . 
百度学术
2. 王宇嘉，贾永刚，牟宏基，孙玉明，郑铎，李思达. 铁路客运服务标准及规章体系构建研究. 铁道运输与经济. 2024(06): 42-47+80 . 百度学术
3. 王宁，柴雪松，李健超，马学志，茅宇琳. 高速铁路无砟轨道典型伤损图谱的设计研究. 铁道建筑. 2023(06): 1-6 . 百度学术
4. 张丁荣，王恪铭，冯心妍. 基于知识图谱和故障树的高速铁路事故致因分析. 铁路计算机应用. 2023(07): 14-18 . 本站查看
5. 孔庆玮，苗茁，冯小芳，杨国柱，龙勋. 基于知识图谱的高速铁路客运调度命令闭环管理方案研究. 铁道运输与经济. 2023(10): 89-98 . 百度学术
6. 郭奕聪，胡维，周政，刘唯佳. 基于知识图谱的铁路企业智能督办系统研究. 铁路计算机应用. 2023(10): 27-31 . 本站查看
其他类型引用(1)
计量
- 文章访问数: 83
- HTML全文浏览量: 24
- PDF下载量: 19
- 被引次数: 7
