基于激光扫描系统的机车车顶异物检测

发布时间：2018-01-08 02:22

本文关键词：基于激光扫描系统的机车车顶异物检测　出处：《华南理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：铁路作为中国重要的基础设施和春运期间最主要的交通工具,其客、货运输系统正处于高速发展阶段,同时也伴随着巨大的运行安全压力。机车车顶异物的存在是列车运行的安全隐患,如损毁列车固有设备,造成列车无法正常运行等。但是机车车顶异物的检测与定位至今仍是一个难点,目前国内外并没有一套完整的体系可以自动完成机车车顶的异物检测。本文提出了一种基于激光扫描结合点云处理的异物检测方法。首先,论文采用一台三维激光扫描仪LMS400对车顶进行扫描。扫描仪不容易受到光线等外界干扰。其次,激光扫描仪的距离传感器获得的数据经速度校正后,在空间坐标系下重构成三维点云。距离传感器获得的数据是极坐标下包括距离与角度的信息,为了获得真实的三维点云,我们在列车前进方向加入速度的因素,使点云成为实际运动过程中的点云。然后,我们使用改进的ICP算法进行点云配准。最后,我们采用基于动态阈值的背景差法来查找异物。本论文在前期用MATLAB进行仿真,实现所有算法并达到检测要求后,将前期仿真的算法转换成C代码。实验仿真结果显示,基于相关熵度量的鲁棒ICP配准算法有较高的配准精度和较快的收敛速度,同时,异物也能准确地被检测出来。论文最后对本方法进行了总结并提出了一些有待改进的地方,对后续的研究任务做出了展望。
[Abstract]:Railway as an important infrastructure and the most important means of transport during the Spring Festival, its passenger and cargo transport system is at a high speed development stage. At the same time, the existence of foreign body on the roof of locomotive is the hidden danger of train operation, such as damage to the inherent equipment of the train. However, the detection and location of foreign bodies on the roof of locomotives is still a difficulty. At present, there is not a complete system can automatically complete the detection of foreign body on locomotive roof. In this paper, a method of foreign body detection based on laser scanning combined point cloud processing is proposed. First of all. In this paper, a 3D laser scanner LMS400 is used to scan the roof. The scanner is not susceptible to external interference such as light. Secondly, the data obtained by the distance sensor of laser scanner is corrected by speed. The data obtained by the distance sensor is the information of distance and angle in polar coordinates. In order to obtain the real 3D point cloud, we add the speed factor in the direction of the train forward. Make the point cloud become the point cloud in the actual motion. Then, we use the improved ICP algorithm for point cloud registration. Finally. We use background difference method based on dynamic threshold to find foreign body. In this paper, we use MATLAB to simulate all the algorithms and achieve the detection requirements. The experimental results show that the robust ICP registration algorithm based on correlation entropy metric has higher registration accuracy and faster convergence speed. Foreign bodies can also be detected accurately. Finally, the method is summarized and some improvements are proposed, and the future research tasks are prospected.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U270.389;TN249

【相似文献】