基于轻量级网络的铁路感兴趣区域异物侵限检测

doi:10.13229/j.cnki.jdxbgxb20210266

摘要/Abstract

摘要：

针对当前基于计算机视觉的铁路异物侵限算法存在错误预警、检测效率低、无法满足轻量级部署等问题，提出了一种基于轻量级网络的铁路感兴趣区域异物侵限检测方法。首先，采用透视变换和三次函数拟合的方法检测铁轨线，通过找到铁轨所在区域，扩展划分出危险区域和安全区域，得到铁路异物侵限检测的感兴趣区域。然后，利用稀疏化和通道剪枝方法对YOLOv3模型进行压缩，构建了轻量级铁路异物检测模型。最后，通过铁路数据集及现场实验进行测试表明，本文方法具有较高的检测精度和检测速度，本文轻量级模型参数空间减小为原有的1/5，检测速度是Faster R-CNN模型的3.4倍，YOLOv3模型的1.3倍，能够快速有效地检测出不同铁路场景危险区域的异物侵限，减少了错误预警。

关键词: 计算机应用, 异物检测, 感兴趣区域划分, 轻量级网络, 铁路

Abstract:

For the problems of false warning， low detection efficiency and insatiable lightweight deployment of railway foreign object intrusion algorithms based on computer vision， a method of detection foreign object intrusion in railway region of interest based on lightweight network is proposed.. Firstly， the railway track line is detected by perspective transformation and cubic function fitting. By finding the area where the rail is located， and then expanding the division of dangerous area and safe area， the region of interest of railway foreign object intrusion detection is obtained. Secondly， a lightweight railway foreign object detection model is constructed by using sparse and channel pruning methods to compress YOLOv3 model. Finally， through the railway data set and field test， it shows that proposed method has high detection accuracy and detection speed. The lightweight model parameter space is reduced to 1/5 of the original， and the detection speed is 3.4 times of Faster R-CNN method， and which is 1.3 times of YOLOv3 method. It can quickly and effectively detect the intrusion of railway foreign object in dangerous areas of different railway scenario， and reduce the error warning.

Key words: computer application, intrusion detection, region of interest division, lightweight network, railway

中图分类号:

TP391.4

陈永,卢晨涛,王镇. 基于轻量级网络的铁路感兴趣区域异物侵限检测[J]. 吉林大学学报(工学版), 2022, 52(10): 2405-2418.

Yong CHEN,Chen-tao LU,Zhen WANG. Detection of foreign object intrusion in railway region of interest based on lightweight network[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(10): 2405-2418.

图/表 18

图1

图2

图3

图4

图5

图6

图7

图8

表1

图9

图10

图11

图12

图13

图14

图15

表2

图16

参考文献 22

1	He D Q, Yao Z K, Jiang Z, et al. Detection of foreign matter on high-speed train underbody based on deep learning[J]. IEEE Access, 2019, 7: 183838-183846.
2	王瑞, 史天运, 包云. 一种基于视频的铁路周界入侵检测智能综合识别技术研究[J].仪器仪表学报, 2020, 41(9): 188-195.
	Wang Rui, Shi Tian-yun, Bao Yun. Research on an intelligent comprehensive recognition technology in railway perimeter intrusion detection based on video[J]. Chinese Journal of Scientific Instrument, 2020, 41(9): 188-195.
3	Simona F, Eduardo F, Francesca D C, et al. Railways track characterization using ground penetrating radar [J].Procedia Engineering, 2016, 143:1193-1200.
4	Dhiraj S, Omaka R. Object drop detection on railway track through rayleigh wave sensing using laser vibrometer[J].IEEE Transactions on Vehicular Technology, 2018, 67(10): 9158-9172.
5	Alexey T, Aleksandrs R, Viktors M. Assessment of cracks in pre-stressed concrete railway sleepers by ultrasonic testing[J]. Procedia Computer Science, 2019, 149: 324-330.
6	Niu H X, Hou T. Fast detection study of foreign object intrusion on railway track[J]. Archives of Transport, 2018, 46(3): 79-89.
7	徐谦, 李颖, 王刚. 基于深度学习的行人和车辆检测[J]. 吉林大学学报: 工学版, 2019, 49(5): 1661-1667.
	Xu Qian, Li Ying, Wang Gang. Pedestrian-vehicle detection based on deep learning[J]. Journal of Jilin University (Engineering and Technology Edition), 2019, 49(5): 1661-1667.
8	Li Y D, Dong H, Li H G,et al. Multi-block SSD based on small object detection for UAV railway scene surveillance[J]. Chinses Journal of Aeronautics, 2020, 33(6): 1747-1755.
9	徐岩, 陶慧青, 虎丽丽. 基于Faster R-CNN网络模型的铁路异物侵限检测算法研究[J]. 铁道学报, 2020, 42(5): 91-98.
	Xu Yan, Tao Hui-qing, Hu Li-li. Railway foreign body intrusion detection based on faster R-CNN network model[J]. Journal of the China Railway Society, 2020, 42(5): 91-98.
10	华夏, 王新晴, 王东, 等. 基于改进SSD的交通大场景多目标检测[J].光学学报, 2018, 38(12): 221-231.
	Hua Xia, Wang Xin-qing, Wang Dong, et al. Multi-objective detection of traffic scene based on improved SSD[J]. Acta Optica Sinica, 2018, 38(12): 221-231.
11	Qi H Y, Xu T H, Wang G, et al. MYOLOv3-Tiny: a new convolutional neural network architecture for real-time detection of track fasteners[J]. Computers in Industry, 2020, 123: No. 103303.
12	王玮, 朱力强. 基于特征图裁剪的高铁周界入侵实时检测算法[J].铁道学报, 2019, 41(9): 74-80.
	Wang Wei, Zhu Li-qiang. Real-time instrusion detection algorithm for high-speed railway based on feature map pruning[J]. Journal of the China Railway Society, 2019, 41(9): 74-80.
13	Daniel S R, Ivan O P, Arturo O L. Automatic region of interest segmentation for breast thermogram image classification[J]. Pattern Recognition Letters, 2020, 135: 72-81.
14	Karakose M, Yaman O, Murat K, et al. A new approach for condition monitoring and detection of rail components and rail track in railway[J]. International Journal of Computational Intelligence Systems, 2018, 11(1): 830-845.
15	Huang Y P, Li Y W, Ci W Y. Lane detection based on inverse perspective transformation and kalman filter[J]. KSII Transactions on Internet and Information Systems, 2018, 12(2): 643-661.
16	. 标准轨距铁路建筑限界 [S].
17	Redmon J, Farhadi A. YOLOv3: an Incremental Improvement[C]∥IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, 2018.
18	Zhang X L, Dong X P, Wei Q J, et al. Real-time object detection algorithm based on improved YOLOv3[J]. Journal of Electronic Imaging, 2019, 28(5): No. 053022.
19	Liu Z, Li J, Shen Z, et al. Learning efficient convolutional networks through network slimming[C]∥Proceedings of 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 2736-2744.
20	李新春, 王藜谚, 王浩童. 基于3DCNN的CSI-cluster室内指纹定位算法[J]. 重庆邮电大学学报:自然科学版, 2020, 32(3): 345-355.
	Li Xin-chun, Wang Li-yan, Wang Hao-tong. CSI-cluster indoor fingerprint localization algorithm based on 3DCNN[J]. Journal of Chongqing University of Posts and Telecommunications (Natural Science Edition), 2020, 32(3): 345-355.
	蔡哲栋, 应娜, 郭春生, 等.YOLOv3剪枝模型的多人姿态估计[J]. 中国图象图形学报, 2021, 26(4): 837-846.
	Cai Zhe-dong, Ying Na, Guo Chun-sheng, et al. Research on multiperson pose estimation combined with YOLOv3 pruning model[J]. Journal of Image and Graphics, 2021, 26(4): 837-846.
22	刘高天, 段锦, 范祺, 等. 基于改进RFBNet算法的遥感图像目标检测[J]. 吉林大学学报:理学版, 2021, 59(5): 1188-1198.
	Liu Gao-tian, Duan Jin, Fan Qi, et al. Target detection for remote sensing image based on improved RFBNet algorithm[J]. Journal of Jilin University (Science Edition), 2021, 59(5): 1188-1198.
	袁梅, 全太锋, 黄俊, 等. 基于卷积神经网络的烟雾检测[J]. 重庆邮电大学学报:自然科学版, 2020, 32(4): 620-629.
	Yuan Mei, Quan Tai-feng, Huang Jun, et al. Smoke detection based on convolutional neural network[J]. Journal of Chongqing University of Posts and Telecommunications (Natural Science Edition), 2020, 32(4): 620-629.

相关文章 15

[1]	周丰丰,朱海洋. 基于三段式特征选择策略的脑电情感识别算法SEE[J]. 吉林大学学报(工学版), 2022, 52(8): 1834-1841.
[2]	白天,徐明蔚,刘思铭,张佶安,王喆. 基于深度神经网络的诉辩文本争议焦点识别[J]. 吉林大学学报(工学版), 2022, 52(8): 1872-1880.
[3]	曲福恒,丁天雨,陆洋,杨勇,胡雅婷. 基于邻域相似性的图像码字快速搜索算法[J]. 吉林大学学报(工学版), 2022, 52(8): 1865-1871.
[4]	赵宏伟,张健荣,朱隽平,李海. 基于对比自监督学习的图像分类框架[J]. 吉林大学学报(工学版), 2022, 52(8): 1850-1856.
[5]	秦贵和,黄俊锋,孙铭会. 基于双手键盘的虚拟现实文本输入[J]. 吉林大学学报(工学版), 2022, 52(8): 1881-1888.
[6]	胡丹,孟新. 基于时变网格的对地观测卫星搜索海上船舶方法[J]. 吉林大学学报(工学版), 2022, 52(8): 1896-1903.
[7]	商拥辉,徐林荣,陈钊锋. 高铁刚性桩⁃筏地基的固结特性及影响因素[J]. 吉林大学学报(工学版), 2022, 52(7): 1588-1597.
[8]	王军,徐彦惠,李莉. 低能耗支持完整性验证的数据融合隐私保护方法[J]. 吉林大学学报(工学版), 2022, 52(7): 1657-1665.
[9]	周丰丰,张亦弛. 基于稀疏自编码器的无监督特征工程算法BioSAE[J]. 吉林大学学报(工学版), 2022, 52(7): 1645-1656.
[10]	康耀龙,冯丽露,张景安,陈富. 基于谱聚类的高维类别属性数据流离群点挖掘算法[J]. 吉林大学学报(工学版), 2022, 52(6): 1422-1427.
[11]	王文军,余银峰. 考虑数据稀疏的知识图谱缺失连接自动补全算法[J]. 吉林大学学报(工学版), 2022, 52(6): 1428-1433.
[12]	陈雪云,贝学宇,姚渠,金鑫. 基于G⁃UNet的多场景行人精确分割与检测[J]. 吉林大学学报(工学版), 2022, 52(4): 925-933.
[13]	方世敏. 基于频繁模式树的多来源数据选择性集成算法[J]. 吉林大学学报(工学版), 2022, 52(4): 885-890.
[14]	李大湘,陈梦思,刘颖. 基于STA⁃LSTM的自发微表情识别算法[J]. 吉林大学学报(工学版), 2022, 52(4): 897-909.
[15]	刘铭,杨雨航,邹松霖,肖志成,张永刚. 增强边缘检测图像算法在多书识别中的应用[J]. 吉林大学学报(工学版), 2022, 52(4): 891-896.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

算法	准确率/%	综合精度/%	检测速度 /（帧·s^-1）
YOLOv3模型	85.43	75.43	45
文献［8］方法	86.38	76.47	26
文献［9］方法	88.32	77.52	17
本文方法	85.41	75.26	58