论文部分内容阅读
动车组运行故障动态图像检测系统(TEDS),通过在轨边安装布置高速线阵采集像机,实现对运行中的列车进行全方位的监控。利用获得的高质量图像,通过机器学习和模式识别的知识,可以实现列车故障的自动化诊断和检测,但是线阵像机拍摄的图像易受动车速度的影响,在图像水平方向上存在几何变形,给后续目标的自动识别和检测带来了困难。为了解决这个问题,设定一组基准图像,对其他时间段所获得的目标图像分别按照对应的基准图像进行配准和重分割,尽量减小列车速度对成像变形的影响。因此,本文结合TEDS系统,利用多分辨率下的图像快速配准方法,实现了后续目标图像的快速分割与对齐。然后,本文提出了一种改进的图像差影技术,通过将对齐之后的目标图像与历史标准图像进行比对分析,快速实现动车故障区域的自动定位和检测。
EMU running fault dynamic image detection system (TEDS), through the rail installation of high-speed line array camera installed to achieve the operation of the train a full range of monitoring. The use of high-quality images obtained through the machine learning and pattern recognition knowledge, can realize the automatic diagnosis and detection of train failure, but the line array camera images are susceptible to the speed of moving vehicles in the image there is a geometric distortion in the horizontal direction, To the follow-up target automatic identification and detection of difficulties. In order to solve this problem, a set of reference images are set, the target images obtained in other time periods are respectively registered and re-divided according to the corresponding reference images, and the influence of the train speed on the image deformation is minimized. Therefore, this paper combined with TEDS system, using rapid image registration method under multi-resolution, to achieve the subsequent target image fast segmentation and alignment. Then, an improved image difference technique is proposed in this paper. By comparing the target image after alignment with the historical standard image, the automatic locating and detecting of the fault area of the vehicle can be quickly realized.