基于3D点云的列车零部件异常检测技术研究
Anomaly Detection of Train Components Based on 3D Point Cloud

随着铁路技术的飞速发展，列车安全问题也备受瞩目。为解决铁路系统快速增加的吞吐量与日益增长的检修压力之间的矛盾，本文提出了一种基于3D点云的列车零部件异常检测方法。首先，建立列车标准点云模板；然后，截取待检部件与相应模板部件进行全局ICP点云配准，利用差异点筛选出异常区域；最后，通过零部件局部检测算法消除误检测区域，得到最终结果。该方法以某地铁车辆段采集数据作为测试集进行了实验，结果表明，算法检测精度高，检测项点覆盖范围较广，具有良好的工业应用价值。
Railway traffic safety has attracted much attention with the rapid development of railway technology. An automatic abnormal railway underbody detection technology based 3D cloud point is proposed to solve the contradiction between the rapidly increasing throughput and the increasing maintenance pressure. Firstly, establishe the train standard point cloud template. Then, the parts to be inspected are intercepted and matched with the corresponding template parts for global ICP point cloud registration, and the abnormal areas are filtered out by using the difference points. Finally, remove the false detection areas by the part local detection algorithm, and the final results are obtained. This method is tested on the dataset collected from a subway depot as a test set. The results show that the algorithm has high detection accuracy, wide coverage of detection items, and good industrial application value.