目前地铁站台门空隙异物检测主流方法存在准确率不高、主观性较强和人力成本高的问题，依靠激光或红外探测方法存在小的物体检测不到、曲型站台有检测死角等问题，地铁列车运行面临重大安全问题。提供地铁站台门空隙光照的灯管长期开启，存在大量电力资源浪费的问题。本文创新性的采用深度学习算法对地铁站台门空隙图像进行了增强，又创新性的采用数字图像处理的方法对地铁站台门空隙异物检测进行了研究，结果表明数字图像处理的方法结果明显优于现阶段主流的人工判断法和激光探测法，能解决地铁列车运作存在的重大安全隐患，减轻地铁列车司机的劳动强度。主要研究如下：

（1）为了减少目标检测系统对站台光照的依赖，节约电力资源成本，本文基于某地铁实际场景，采用置顶式近距离全方位的采集地铁站台门空隙图像构建数据集，研究了基于Zero-DCE的地铁站台门空隙图像增强算法，并对Zero-DCE进行改进，实验结果表明：经改进的Zero-DCE算法增强的地铁站台门空隙图像标准差较原图增加了33.7%，平均梯度提升59%，信息熵提高8%。

（2）采用HOG算法对地铁站台门空隙异物图像进行特征提取，使用SVM算法进行分类，预测结果检测率为0.85，误检率为0.14。可以在一定程度上解决电客车司机确认地铁站台门空隙安全的问题。

（3）采用深度学习YOLOv8算法对地铁站台门空隙异物进行检测，通过添加Ghost模块和小目标检测层对原算法进行改进，实验结果表明Precision可达0.94，Recall达到0.89，mAP@0.5达到0.95，并在低照度模式下也具有较好的性能指标，很好的解决地铁列车站台门空隙异物检测难的问题。本课题的研究创新性的解决现阶段地铁运营中站台门空隙异物检测难问题的同时，也为逐步推进的地铁无人驾驶技术提供了技术支持，因此具有重大应用前景。

At present,the mainstream method of foreign body detection in the gap of subwaystations has problems such as not high accuracy,strong subjectivity and high labor cost.Thereare problems such as small objects and dead detection in curved platform,and the subwaytrain operation is faced with major safety problems.The lamp that provides the gap of thesubway station gate is open for a long time,and there is a lot of waste of power resources.Thispaper adopts the deep learning algorithm to enhance the subway gate gap image,and thedigital image processing method,and the results show that the digital image processingmethod is significantly better than the current mainstream artificial judgment method and laserdetection method,which can solve the major safety hidden danger in subway train operation,and reduce the labor intensity of subway train drivers.The main studies are conducted asfollows:

(1)In order to reduce the dependence of target detection system on platform lighting andsave the cost of power resources,this paper based on the actual scenario of the subway stationbased on Zero-DCE,and improves the Zero-DCE.The experimental results show that theimproved Zero-DCE algorithm enhances the standard deviation increased by 33.7%comparedwith the original image,the average gradient increased by 59%,and the information entropyincreased by 8%.

(2)HOG algorithm was used to extract the features of the subway gate gap foreign bodyimage,and SVM algorithm was used to classify.The detection rate of the prediction result was0.85 and the misdetection rate was 0.14.To a certain extent,it can solve the problem ofconfirming the safety of the subway station gate gap.

(3)Using deep learning YOLOv8 algorithm to subway station gate gap foreign bodydetection,by adding Ghost module and small target detection layer to improve the originalalgorithm,the experimental results show that the Precision can reach 0.94,Recall reached 0.89,mAP@0.5 reached 0.95,and in low illumination mode also has good performance index,verygood solve the subway train platform door gap foreign body detection problem.The research of this topic not only solves the problem of foreign body detection in thesubway operation at the present stage,but also provides technical support for the graduallypromoted subway driverless technology,so it has a great application prospect.

王福文,梁帅文,冯爱军.2023年我国城市轨道交通数据统计与发展分析[J].隧道建设,2024,44(02):393-400.

李爱东, 宋剑伟, 袁伟. 深圳市智慧城轨发展实践及趋势[J]. 现代城市轨道交通, 2020(12): 9-15.

中国城市轨道交通协会. 中国城市轨道交通智慧城轨发展纲要[J]. 城市轨道交通, 2020(4): 8-23.

T/CAMET 11001.1— 2019. 智慧城市轨道交通信息技术架构及网络安全规范第l部分:总体需求[S]. 北京: 中国铁道出版社, 2019:7-27．

T/CAMET 11001.2— 2019. 智慧城市轨道交通信息技术架构及网络安全规范第2部分:总体需求[S]. 北京: 中国铁道出版社, 2019:8-35.

刘伟铭,李冠金,汪良,等. 轨道交通站台门与列车间隙异物的四目定位方法研究[J]. 铁道标准设计,2023,67(8):147-152.

侯妍君,李鑫磊,申樟虹,等. 站台门与车门间隙探测装置研究[J]. 现代机械,2020(1):94-97.

郭禹. 激光雷达技术在北京地铁站台门间隙探测系统的应用[J]. 中国科技纵横,2019(10):67-69.

钱广民,阎娟. 全自动运行线路间隙探测装置与行车相关系统联动方式优化方案[J]. 铁路计算机应用,2022,31(8):77-80.

TAN FEIGANG, ZHAO XIAOJU, LIAO QUANMI. A Robust Detection of Foreign Matter between Train Doors and Psd of Subway Station[C]. //2020 12th International Conference on Measuring Technology and Mechatronics Automation: 12th International Conference on Measuring Technology and Mechatronics Automation (ICMTMA 2020), 28-29 Feb. 2020, Phuket, Thailand.:Institute of Electrical and Electronics Engineers, 2020:246-248.

雷焕宇,刘伟铭. 基于K-means算法的地铁站台异物检测[J]. 计算机与现代化,2018(6):42-46.

严华兵. 基于BP神经网络的列车站台门异物检测[J]. 中国新技术新产品,2023(16):86-88.

谭飞刚,翟聪. 地铁曲线站台空隙异物自动检测预警系统研究[J]. 铁路计算机应用,2018,27(6):59-61.

施伟,郑兆鹏,班琪. 基于改进Hough算法的地铁屏蔽门异物检测技术[J]. 城市轨道交通研究,2016,19(11):81-84.

饶美婉. 新型地铁站台门与车门之间顶置式防夹人自动探测系统的设计[J]. 城市轨道交通研究,2017,20(5):143-146.

张嘉超,张伟.基于特征的地铁门与安全门间隙异物检测方法[J].南京工程学院学报:自然科学版, 2022, 20(4):35-40.

TAN FEIGANG, ZHAO XIAOJU, LIU KAIYUAN. Foreign body detection between train doors and PSD based on multi-source data fusion[C]. //2020 13th International Conference on Intelligent Computation Technology and Automation: ICICTA 2020, Xi an, China, 24-25 October 2020.:Institute of Electrical and Electronics Engineers, 2020:596-597.

Cheng F, Cai H. Deep learning-based object detection between train and rail transit platform door[J]. International Journal of Grid and Utility Computing, 2022,13(5): 526-537.

杨嘉能,李华,田宸玮,等.基于自适应校正的动态直方图均衡算法[J].计算机工程与设计,2021,42(05):1264-1270.

闫哲,蒋砾,杨帆,等.基于双直方图均衡算法的红外图像增强[J].红外技术,2022,44(09):944-950.

Gu Z H, Li F, Fang F M, et al. A novel retinex-based frac-tional-order variational model for images with severely low light[J]. IEEE Transactions on Image Processing, 2019, 29: 3239-3253.

Hao S J, Han X, Guo Y R, et al.Low-light image enhancement with semi-decoupled decomposition[J]. IEEE Transactions on Multimedia, 2020, 22(12): 3025-3038.

刘家朋,赵宇明,胡福乔.基于单尺度Retinex算法的非线性图像增强算法[J].上海交通大学学报,2007,(05):685-688.

魏然然,詹伟达,朱德鹏,等.改进多尺度的Retinex红外图像增强[J].液晶与显示,2021,36(03):465-474.

蔡云.带色彩恢复的多尺度Retinex改进型算法研究[J].宁夏师范学院学报,2022,43(10):67-73+100.

Lore K G,Akintayo A,Sarkar S.LLNet:a deep autoencoder approach to natural low-light image enhancement[J]. Pattern Recognition, 2017, 61: 650-662 .

Fan M H,Wang W J,Yang W H,et al.Integrating semantic segmentation and retinex model for low-light image enhance-ment[C]//Proceedings of the 28th ACM International Confer-ence on Multimedia. New York: ACM Press, 2020: 2317-2325.

Lv F F,Lu F,Wu J H,et al.MBLLEN: low-light image/video enhancement using CNNs[C] //Proceedings of the British Ma-chine Vision Conference. Swansea:BMVC Press,2018: 220(1): 4.

Jiang Y F,Gong X Y, Liu D,et al.EnlightenGAN:deep light enhancement without paired supervision[J]. IEEE Transactions on Image Processing, 2021,30: 2340-2349.

Guo C L,Li C Y,Guo J C,et al.Zero-reference deep curve estimation for low-light image enhancement[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Los Alamitos: IEEE Computer Society Press, 2020: 1777-1786.

刘林云,陈开颜,李雄伟,张阳,谢方方.基于卷积神经网络的旁路密码分析综述[J/OL].计算机科学:1-12[2022-03-28]

Li Y D,Hao Z B,Lei H.Survey of convolutional neural network[J].Journal of Computer Applications,2016,36(9):2508-2515.

陶建敏.基于深度学习的车牌识别技术的研究与实现[D].广西师范大学,2020.DOI:10.27036/d.cnki.ggxsu.2020.000487.

饶宸.基于深度学习的光学遥感目标检测技术研究[D].东华理工大学,2020.DOI:10.27145/d.cnki.ghddc.2020.000153.

Zhou W, Gao S, Zhang L, et al. Histogram of Oriented Gradients Feature Extraction from Raw Bayer Pattern Images[J]. IEEE Transactions on Circuits and Systems II: Express Briefs, 2020, 67(5): 946-950.

Xiao Z ,Song Q ,Wei Y , et al.Use of survival support vector machine combined with random survival forest to predict the survival of nasopharyngeal carcinoma patients.[J].Translational cancer research,2023,12(12):3581-3590.

Moradi S，Zayed T，Nasiri F，et a1．Automated anomaly detection and localization in sewer inspection videos using proportional data modeling and deep leaming-based text recognition[J]．Journal of Infrastructure Systems，2020,26(3)：04020018．

马利伟,周明阳,刘国宁.高铁建设环境人员防入侵安全检测方法研究[J].机械设计与制造,2022,(02):182-187.

姚群力,胡显,雷宏.深度卷积神经网络在目标检测中的研究进展[J].计算机工程与应用.2018.54(17): 1-9.

Shilpa R ,Deepika G ,Sandeep K .Object detection and recognition using contour based edge detection and fast R-CNN[J].Multimedia Tools and Applications,2022,81(29):42183-42207.

Christine D ,Henoch C J .Combination of Deep Cross-Stage Partial Network and Spatial Pyramid Pooling for Automatic Hand Detection[J].Big Data and Cognitive Computing,2022,6(3):85-85.

Girshick R,Fast R-CNN [C]//IEEE International Conference on Computer Vision (ICCV),Santiago Chile,December7-13,2015, New York: IEEE,2015:1440-1448.

Sunardi ,Yudhana A ,Putri W R A .Optimization of Breast Cancer Classification Using Faster R-CNN[J].Revue d'Intelligence Artificielle,2023,37(1):

包晓敏,王思琪.基于深度学习的目标检测算法综述[J].传感器与微系统,2022,41(04):5-9.

姜竣,翟东海.基于空洞卷积与特征增强的单阶段目标检测算法[J].计算机工程,2021,47(07):232-238+248.

Redmon J, Divvala S,R.Girshick and A.Farhadi,"You Only Look Once: Unified,Real-Time Object Detection," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR),2016,pp.779-788.

JOSEPH REDMON, ALI FARHADI. YOLO9000: Better, Faster, Stronger[C]. //2017 IEEE Conference on Computer Vision and Pattern Recognition: CVPR 2017, Honolulu, Hawaii, USA, 21-26 July 2017, pages 1-751, [v.1].:Institute of Electrical and Electronics Engineers, 2017:6517-6525.

Mirye K ,Youngmin K .Parcel Classification and Positioning of Intelligent Parcel Storage System Based on YOLOv5[J].Applied Sciences,2022,13(1):437-437.

S. B V ,S. S ,A. S S , et al.Autonomous face mask detection using single shot multibox detector, and ResNet-50 with identity retrieval through face matching using deep siamese neural network[J].Journal of Ambient Intelligence and Humanized Computing,2023,14(8):11195-11205.

KAI HAN, YUNHE WANG, QI TIAN, et al. GhostNet: More Features From Cheap Operations[C]. //2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020), 13-19 June 2020, Seattle, WA, USA.:Institute of Electrical and Electronics Engineers, 2020:1577-1586.

赵佳英.超参数设置对图像分类技术的影响[J].科技与创新,2022(14):160-162+165.