A virtual rail train path recognition method based on binocular stereo vision

ZENG Junwei; GENG Li; LU Chenyang; REN Lihui

doi:10.13890/j.issn.1000-128X.2024.01.122

您当前的位置：

首页 >

文章列表页 >

A virtual rail train path recognition method based on binocular stereo vision

Intelligent Technology | 更新时间：2024-08-02

- A virtual rail train path recognition method based on binocular stereo vision
- Electric Drive for Locomotives Issue 1, Pages: 145-151(2024)
- 作者机构：
  
  同济大学铁道与城市轨道交通研究院，上海 201804
- 作者简介：
- 基金信息：
- DOI：10.13890/j.issn.1000-128X.2024.01.122
  CLC： U239.5
- Published：10 January 2024，
  
  Received：23 August 2021，
  
  Revised：29 December 2023，
- 稿件说明：
扫描看全文
曾俊玮, 耿力, 卢晨旸, 等. 一种基于双目立体视觉的虚拟轨道列车路径识别方法[J]. 机车电传动, 2024(1): 145-151.

ZENG Junwei, GENG Li, LU Chenyang, et al. A virtual track train path recognition method based on binocular stereo vision[J]. Electric drive for locomotives, 2024(1): 145-151.
曾俊玮, 耿力, 卢晨旸, 等. 一种基于双目立体视觉的虚拟轨道列车路径识别方法[J]. 机车电传动, 2024(1): 145-151. DOI：10.13890/j.issn.1000-128X.2024.01.122.

ZENG Junwei, GENG Li, LU Chenyang, et al. A virtual track train path recognition method based on binocular stereo vision[J]. Electric drive for locomotives, 2024(1): 145-151. DOI：10.13890/j.issn.1000-128X.2024.01.122.

摘要

为了克服单目视觉的局限性和提高虚拟轨道列车的路径感知能力，文章提出一种基于双目立体视觉的虚拟轨道列车路径识别方法。针对虚拟轨道列车车道块明显、外形完整、形状独特等特点，提出了基于Mask R-CNN的车道检测模型。同时，针对通用的双目匹配算法计算量大和对重复物体匹配较难的缺点，提出了融合多目标跟踪的双目匹配算法。该方法通过左右相机各自检测车道块，进行多目标跟踪来分配ID，对左右图像中的各车道块实现有序、定向的双目匹配，并重建出车辆坐标系下路径的三维坐标，有效提升了车辆系统路径感知能力，为其循迹控制、自主定位、相对位姿估计等提供了更加直接准确的输入信息。试验结果表明，本文方法对路径的三维重建具有较高的准确性，并且在不同路况场景下具有较强的适应性。

Abstract

This paper presents a path recognition method based on binocular stereo vision for virtual rail train

to overcome the limitations of monocular vision and improve the path perception ability of virtual rail train. To achieve this

a lane detection model based on Mask R-CNN was developed

incorporating their characteristics such as obvious lane blocks

contour integrity

and unique shape. Meanwhile

a binocular matching algorithm based on multi-target tracking was proposed

addressing the drawbacks of general binocular matching algorithms including substantial computations and deficient ability to match repeated objects. The proposed method incorporated both left and right cameras to detect lane blocks respectively

and enabled the assignment of IDs through multi-target tracking. There features allowed for an orderly and directional binocular matching of lane blocks in the left and right images

and the reconstruction of three-dimensional coordinates of the path in the tram coordinate system. This method was demonstrated effective in improving the path perception ability of trams and providing more direct and accurate input information to enable various functions such as tracking control

autonomous positioning and relative pose estimation of trams. Additionally

experimental results show the high accuracy of the proposed method in 3D path reconstruction and its strong adaptability to different road conditions.

关键词

虚拟轨道列车路径识别双目立体视觉目标检测多目标跟踪双目匹配三维重建

Keywords

virtual rail trainlane recognitionbinocular stereo visiontarget detectionmulti-target trackingbinocular matching3D reconstruction

references

陆海英, 丁铁成, 杨岗. 智能轨道列车综述[J]. 电子世界, 2021(3): 166-167.

LU Haiying, DING Tiecheng, YANG Gang. Overview of intelligent rail trains[J]. Electronics world, 2021(3): 166-167.

BORKAR A, HAYES M, SMITH M T. A novel lane detection system with efficient ground truth generation[J]. IEEE transactions on intelligent transportation systems, 2012, 13(1): 365-374.

CAI Hao, HU Zhaozheng, HUANG Gang, et al. Robust road lane detection from shape and color feature fusion for vehicle self-localization[C]//IEEE. 2017 4th International Conference on Transportation Information and Safety (ICTIS). Banff: IEEE, 2017: 1009-1014.

WANG Jianzhuang, CHEN Youping, XIE Jingming, et al. Model-based lane detection and lane following for intelligent vehicles[C]//IEEE. 2010 Second International Conference on Intelligent Human-Machine Systems and Cybernetics. Nanjing: IEEE, 2010: 170-175.

LI Jun, MEI Xue, PROKHOROV D, et al. Deep neural network for structural prediction and lane detection in traffic scene[J]. IEEE transactions on neural networks and learning systems, 2017, 28(3): 690-703.

HUVAL B, WANG Tao, TANDON S, et al. An empirical evaluation of deep learning on highway driving[EB/OL]. (2015-04-17) [2023-10-09]. https://arxiv.org/abs/1504.01716https://arxiv.org/abs/1504.01716.

PAN Xingang, SHI Jianping, LUO Ping, et al. Spatial as deep: spatial CNN for traffic scene understanding[EB/OL]. (2017-12-17) [2023-10-09]. https://arxiv.org/abs/1712.06080https://arxiv.org/abs/1712.06080.

王小坤, 贾庆轩, 谭胜. 一种单目视觉测距方法的研究[J]. 科技信息(学术版), 2008(34): 96.

WANG Xiaokun, JIA Qingxuan, TAN Sheng. Research on a monocular vision ranging method[J]. Science & technology information, 2008(34): 96.

李彪. 基于双目立体视觉三维重建技术研究[D]. 西安: 西安科技大学, 2018.

LI Biao. Research on 3D reconstruction technology based on binocular stereo vision[D]. Xi'an: Xi'an University of Science and Technology, 2018.

白明, 庄严, 王伟. 双目立体匹配算法的研究与进展[J]. 控制与决策, 2008, 23(7): 721-729.

BAI Ming, ZHUANG Yan, WANG Wei. Progress in binocular stereo matching algorithms[J]. Control and decision, 2008, 23(7): 721-729.

贺捷. 智能轨道快运系统株洲智轨体验线项目特点及适用性分析[J]. 城市轨道交通研究, 2020, 23(1): 27-30.

HE Jie. Characteristics of Zhuzhou intelligent rail express line project and feasibility analysis[J]. Urban mass transit, 2020, 23(1): 27-30.

翁木云, 谢宇昕. 一种改进的自适应质心跟踪算法[J]. 空军工程大学学报(自然科学版), 2009, 10(2): 81-85.

WENG Muyun, XIE Yuxin. An improved adaptive centroid tracking algorithm[J]. Journal of air force engineering university (natural science edition), 2009, 10(2): 81-85.

曾俊玮, 季元进, 任利惠, 等. 卡尔曼滤波器与神经网络串行的轮胎载荷识别模型[J]. 振动与冲击, 2023, 42(11): 262-270.

ZENG Junwei, JI Yuanjin, REN Lihui, et al. A serial tire load identification model based on Kalman filter and neural network[J]. Journal of vibration and shock, 2023, 42(11): 262-270.

孙明珠. 面向低纹理区域的局部立体匹配算法研究[D]. 秦皇岛: 燕山大学, 2018.

SUN Mingzhu. Research on local stereo matching algorithm for low texture area[D]. Qinhuangdao: Yanshan University, 2018.

Views

下载量

CSCD

CNKI被引量

Alert me when the article has been cited

提交

Tools

Publicity Resources

Arcing detection method of metro pantograph based on improved yolov4-tiny

Research on running system of a 100% low floor virtual rail train

Research on lidar target detection for metro scene

Related Author

AI Shengyong

DING Jianming

ZHANG Qingsong

Wen LI

Haiying LU

Lihui REN

Chunyou GAO

Nuo LI

Related Institution

State Key Laboratory of Rail Transit Vehicle System, Southwest Jiaotong University

CRRC Changchun Railway Vehicles Co., Ltd.

Railway and Urban Rail Transit Research Institute, Tongji University

CRRC Zhuzhou Institute Co., Ltd.

The State Key Laboratory of Advanced;Design and Manufacturing for Vehicle Body, Hunan University

⁰