基于机器视觉的轨道交通自动测距研究

doi:10.16180/j.cnki.issn1007-7820.2022.09.006

摘要/Abstract

摘要：

行驶中的两辆列车之间保持安全距离是避免列车追尾事故发生的重要条件。由于机器视觉获得的图像数据信息丰富,可以根据采集到的图像进行多方面的集成检测,所以文中提出了一种基于机器视觉的列车测距方法。该方法以列车两条轨道不变的间距(1 435 mm)作为基准来推算列车之间距离。利用卷积神经网络对单目相机采集到的图像进行处理和分析,提取所需的轨道特征,再基于已有的小孔成像原理推导出世界坐标系与像素坐标系之间的映射关系,从而优化列车之间距离的计算式。实验结果表明,测距系统的误差率<6%,并且系统测量时间在40 ms之内,说明该方法实现了将测距与在图像中获取其他信息的有效融合与集成,可用于对列车制动距离进行判断。

关键词: 距离测量, 机器视觉, 单目相机, 深度学习, 图像处理, 轨道检测, 小孔成像原理, 卷积神经网络

Abstract:

A safe distance between two moving trains is an important condition to avoid train rear-end collision. Since the image data obtained by machine vision is rich in information and can be integrated in many aspects based on the collected images, this study proposes a distance measurement method based on monocular machine vision. This method uses the constant distance between the two tracks of the train (1 435 mm) as a benchmark to estimate the distance between trains. The images collected by the monocular camera are processed and analyzed by the convolution neural network to extract the track features. Based on the existing small hole imaging principle, the mapping relationship between the world coordinate system and the pixel coordinate system is derived, so as to optimize the calculation formula of the distance between trains. The experimental results show that the error rate of the system is less than 6%, and the measurement time of the system is within 40 ms, indicating that the method realizes the effective fusion and integration of ranging and other information obtained in the image, and can be used to judge the braking distance of the train.

Key words: distance measurement, machine vision, monocular camera, deep learning, image processing, track detection, pinhole imaging model, convolution neural network

中图分类号:

TP391

毕嘉桢,沈拓,张轩雄. 基于机器视觉的轨道交通自动测距研究[J]. 电子科技, 2022, 35(9): 37-43.

BI Jiazhen,SHEN Tuo,ZHANG Xuanxiong. A Research on Distance Measurement Between Trains in Rail Transit Based on Machine Vision[J]. Electronic Science and Technology, 2022, 35(9): 37-43.

图/表 8

图1

图2

图3

图4

图5

图6

表1

图7

参考文献 18

[1]	Jung H G, Cho Y H, Yoon P J, et al. Scanning laser radar-based target position designation for parking aid system[J]. IEEE Transactions on Intelligent Transportation Systems, 2008, 9(3):406-424. doi: 10.1109/TITS.2008.922980
[2]	Ibisch A, Stefan S, Harald A, et al. Towards autonomous driving in a parking garage: Vehicle localization and tracking using environment-embedded LIDAR sensors[C]. Gold Coast: Proceedings of IEEE Intelligent Vehicles Symposium, 2013.
[3]	Kenue S K. Selection of range and azimuth angle parameters for a forward looking collision warning radar sensor[C]Detroit:Proceedings of the Intelligent Vehicles'95,Symposium,1995.
[4]	Sugimoto S, Tateda H, Takahashi H, et al. Obstacle detection using millimeter-wave radar and its visualization on image sequence[C]. Cambridge: Proceedings of the Seventeenth International Conference on Pattern Recognition, 2004.
[5]	Qin H, Lan H, Zhang X.Adjusting method for reachable space during automatic parking comprises e.g. starting parking when virtual obstacle on expandable boundary of reachable space to enter blind zone range of ultrasonic ranging radar ranging result is obtained:China, CN111198376-A[P].[2020-05-26].
[6]	Suhr J K, Jung H G. Automatic parking space detection and tracking for underground and indoor environments[J]. IEEE Transactions on Industrial Electronics, 2016, 63(9):5687-5698. doi: 10.1109/TIE.2016.2558480
[7]	Mockel S, Scherer F, Schuster P F. Multi-sensor obstacle detection on railway tracks[C]. Columbus: Proceedings of the IEEE Intelligent Vehicles Symposium, 2003.
[8]	Bertozzi M, Broggi A, Fascioli A. Vision-based intelligent vehicles: State of the art and perspectives[J]. Robotics and Autonomous Systems, 2000, 32(1):1-16. doi: 10.1016/S0921-8890(99)00125-6
[9]	Olaverri-Monreal C, Krizek G C, Michaeler F, et al. Collaborative approach for a safe driving distance using stereoscopic image processing[J]. Future Generation Computer Systems-The International Journal of Escience, 2019, 95(2):880-889.
[10]	Chang T H, Chou C J. Rear-end collision warning system on account of a rear-end monitoring camera[C]. Xi'an: Proceedings of IEEE Intelligent Vehicles Symposium, 2009.
[11]	Wang Z Y, Wu X K, Yu G Z, et al. Efficient rail area detection using convolutional neural network[J]. IEEE Access, 2018(6):77656-77664.
[12]	Wang Z L, Cai B G, Jia C X, et al. Geometry constraints-based visual rail track extraction[C]. Guilin:Proceedings of the Twelfth World Congress on Intelligent Control and Automation, 2016.
[13]	朱斌, 刘子龙. 基于新型初始模块的卷积神经网络图像分类方法[J]. 电子科技, 2021, 34(2):52-56.
	Zhu Bin, Liu Zilong. Image classification method using convolutional neural network based on new initial module[J]. Electronic Science and Technology, 2021, 34(2):52-56.
[14]	程俊华, 曾国辉, 刘瑾. 基于深度学习的复杂背景图像分类方法研究[J]. 电子科技, 2020, 33(12):59-66.
	Cheng Junhua, Zeng Guohui, Liu Jin. Research on complex background image classification method based on deep learning[J]. Electronic Science and Technology, 2020, 33(12):59-66.
[15]	马颂德, 张正友. 计算机视觉:计算理论与算法基础[M]. 北京: 科学出版社,1997.
	Ma Songde, Zhang Zhengyou. Computer vision: fundamentals of computational theory and algorithms[M]. Beijing: Science Press,1997.
[16]	徐德, 谭民, 李原. 机器人视觉测量与控制[M]. 北京: 国防工业出版社, 2008.
	Xu De, Tan Min, Li Yuan. Robot vision measurement and control[M]. Beijing: National Defense Industry Press, 2008.
[17]	Zhang Z Y. A flexible new technique for camera calibration[J]. IEEE Transactions on Pattern Analysis And Machine Intelligence, 2000, 22(11):1330-1334. doi: 10.1109/34.888718
[18]	Zhang Z Y. Camera calibration with one-dimensional objects[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(7):892-899. doi: 10.1109/TPAMI.2004.21

实际距离/m	测量结果/m	测量耗时/ms	错误率/%
40	38.2	35	4.50
60	57.5	36	4.17
80	77.2	32	3.88
100	97.1	37	2.90
120	116.5	34	2.92
140	134.5	35	3.86
160	153.1	38	4.31
180	170.7	34	5.17
200	188.2	31	5.90