基于加速区域卷积神经网络的高铁接触网承力索底座裂纹检测研究  被引量:8

Crack Detection of Messenger Wire Supporter in Catenary Support Devices of High-speed Railway Based on Faster R-CNN

在线阅读下载全文

作  者:刘凯[1] 刘志刚[1] 陈隽文 LIU Kai;LIU Zhigang;CHEN Junwen(School of Electrical Engineering,Southwest Jiaotong University, Chengdu 610031, China)

机构地区:[1]西南交通大学电气工程学院

出  处:《铁道学报》2019年第7期43-49,共7页Journal of the China Railway Society

基  金:国家自然科学基金(U1734202);四川省青年科技创新研究团队专项计划(2016TD0012)

摘  要:针对高速铁路接触网支撑结构中承力索底座裂纹的问题,提出一种利用加速区域卷积神经网络与Beamlet变换相结合的图像检测方法。该方法使用加速区域卷积神经网络实现对承力索底座在待检测图像中的识别定位,然后根据定位的承力索底座图像特点,通过Radon变换等预处理操作对承力索底座疑似裂纹区域精确定位,最后使用基于Beamlet变换的局部链搜索算法快速得到裂纹信息,实现承力索底座裂纹故障的可靠诊断。实验表明:该方法能在复杂的接触网支撑与悬挂装置图像中准确定位识别承力索底座裂纹故障,对拍摄距离、拍摄角度以及曝光度等因素具有很好的适应性,且具有较高的检测效率。A vision-based method to detect the cracks of messenger supporters of high-speed railway catenary is proposed, which is based on the faster region-based convolutional neural networks (Faster R-CNN) and the Beamlet transform.First, Faster R-CNN is adopted to recognize and extract messenger wire supporters.Then, based on the characteristics of messenger wire supporters, the Radon transform is applied to precisely extract the suspected crack areas of the messenger wire supporters.The multi-scale feature information is extracted from suspected crack areas using the Beamlet transform, the local curve search algorithm is used to extract the cracks.Experiments show that the proposed method can efficiently identify the crack failure of messenger wire supporters in complex catenary images, and has a good adaptability to the different shooting distances, shooting angles and illumination intensity.This method can greatly enhance the detection efficiency with a high accuracy.

关 键 词:高铁接触网 承力索底座 加速区域卷积神经网络 BEAMLET变换 

分 类 号:U225.47[交通运输工程—道路与铁道工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象