交通运输系统工程与信息

交通运输系统工程与信息 ›› 2012, Vol. 12 ›› Issue (4): 79-83.

• 智能交通系统与信息技术 • 上一篇    下一篇

基于车载前视摄像机的轨道异物检测

同磊a,b，朱力强*a,b，余祖俊b，郭保青b   

  1. 北京交通大学 a轨道交通控制与安全国家重点实验室; b机械与电子控制工程学院, 北京 100044
  • 收稿日期:2012-03-21 修回日期:2012-06-06 出版日期:2012-08-25 发布日期:2012-09-07
  • 作者简介: 同磊（1987- ），男，陕西合阳人，硕士生.
  • 基金资助:

    轨道交通控制与安全国家重点实验室自主研究课题（RCS2009ZT012）;国家863计划项目（2011AA11A102）.

Railway Obstacle Detection using Onboard Forward-Viewing Camera

TONG Lei a,b,ZHU Li-qiang a,b,YU Zu-junb,GUO Bao-qingb   

  1. a. State Key Laboratory of Rail Traffic Control and Safety; b. School of Mechanical,Electronic and Contral engineering, Beijing Jiaotong University, Beijing 100044, China
  • Received:2012-03-21 Revised:2012-06-06 Online:2012-08-25 Published:2012-09-07

PDF

被引次数

20

摘要：

轨道交通线路净空安全是确保列车平稳、不间断运行的基础.由于轨间异物对行车安全产生严重影响，所以基于非轨道电路的轨道异物入侵检测系统在铁路系统中具有十分广泛的应用前景.本文提出了一种基于移动车载摄像机检测轨间异物的方法.首先，通过钢轨识别算法自动定位钢轨位置，并确定列车前方轨道是否有其它列车或公路车辆等大型异物，若有则进行报警；之后，基于边缘检测的异物检测算法自动检测轨间可疑小异物，同时提取可疑异物的尺度信息和颜色索引参数等相关特征；最后，用支持向量机（SVM）来对可疑小异物区域进行分类和辨识.车载实验结果表明，该方法可以有效地检测出轨间异物.

关键词: 智能交通, 异物检测, 机器视觉, 支持向量机

Abstract:

Track clearance is the foundation of the safe and continuous operation of railway system. Nontrack circuitbased intrusion and obstacle detection techniques may be well applied for railway systems because the rail obstacles always greatly reduced the train speed. In this paper, a machine visionbased obstacle detection method is proposed based on the onboard forwardviewing camera and realtime image processing algorithms. First, a rail recognition algorithm is developed to automatically locate the rails in front of the train and find whether the obstacles in track segment are large enough to block the whole track ahead, such as, rail or road vehicles. Then, a small obstacle detection algorithm based on edge detection is used to detect the candidate obstacle areas. Typical features of the obstacle areas, such as dimension and color index parameters, are thus obtained. Finally, a support vector machine(SVM) is trained to classify the candidate obstacle areas and verify the true obstacles. Experiments show that the proposed method can effectively detect obstacles.

Key words: intelligent transportation, obstacle detection, machine vision, support vector machines

中图分类号: