基于机器视觉的铁路道口异物检测方法研究

发布时间：2018-01-30 13:49

本文关键词： 机器视觉铁路道口异物检测三维重建 Mean-Shift算法　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着机器视觉理论研究的不断快速发展,机器视觉已逐渐应用到各行业领域。铁路道口作为道路与铁路平面相交处,异物侵界引发的交通事故时常发生,带来不必要的人员伤亡和经济损失。对此,本文给出了一种基于机器视觉的铁路道口异物检测方法,有利于铁路道口行车安全。针对双目视觉检测系统中的摄像机参数标定、图像校正、立体匹配、三维重建、异物定位与跟踪等方面问题进行了研究分析,并给出了一些改进的方法,取得了较好的效果。本文的主要工作如下:(1)在分析现有铁路道口异物检测方法的基础上,设计了一种适用于异物检测系统的总体方案,给出了系统的结构功能描述。(2)针对传统的侵界异物定位精度低的问题,给出了一种基于半全局匹配(SGM)算法的双目视觉三维重建的方法,经实验测试,对于离相机对中心7米范围内的异物,提取的深度信息与实际距离之间的相对误差低于3%,改善了传统基于绝对值差和(SAD)算法的三维重建结果不准的问题。(3)针对传统的异物检测方法误报率高的问题,给出了一种双目视觉系统与Mean-Shift跟踪技术相结合的异物检测方法,实现了一种新的双目异物检测系统。其通过双摄像机的跟踪定位改善了运动目标定位不准等缺陷,实验结果显示,该系统能在一定时间内检测异物的运动状态,克服了传统方法容易误判断的问题。
[Abstract]:With the rapid development of machine vision theory, machine vision has been gradually applied to various fields. As the intersection of road and railway plane, traffic accidents caused by foreign body invasion often occur. It brings unnecessary casualties and economic losses. In this paper, a method of detecting foreign bodies in railway crossings based on machine vision is presented. In this paper, the camera parameter calibration, image correction, stereo matching, 3D reconstruction, foreign body location and tracking in binocular vision detection system are studied and analyzed. Some improved methods are given and good results are obtained. The main work of this paper is as follows: 1) on the basis of analyzing the existing methods of detecting foreign bodies at railway crossings. A general scheme for foreign body detection system is designed, and the structure and function description of the system is given. This paper presents a method of 3D reconstruction of binocular vision based on semi-global matching SGM algorithm. The experimental results show that the foreign body in the center of the camera is within 7 meters from the center of the camera. The relative error between the extracted depth information and the actual distance is less than 3%. Improved the traditional algorithm based on absolute value difference and SAD algorithm of 3D reconstruction results of inaccurate problem. (3) in view of the traditional foreign body detection method of high false alarm rate problem. A foreign body detection method combining binocular vision system with Mean-Shift tracking technology is presented. A new binocular foreign body detection system is implemented, which can improve the imprecision of moving object location by tracking and positioning with two cameras. The experimental results show that the system can detect the moving state of foreign body in a certain time. It overcomes the problem that the traditional method is easy to misjudge.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U298;TP391.41

【参考文献】