Intelligent Rail Flaw Detection System Based on Deep Learning and Support Vector Machine

Jiangping LUO; Xizhuo YU; Jingwei CAO; Weihong DU

doi:10.13890/j.issn.1000-128x.2021.02.016

您当前的位置：

首页 >

文章列表页 >

Intelligent Rail Flaw Detection System Based on Deep Learning and Support Vector Machine

Intelligent Technology | 更新时间：2021-12-09

- Intelligent Rail Flaw Detection System Based on Deep Learning and Support Vector Machine
- Electric Drive for Locomotives Issue 2, Pages: 100-107(2021)
- 作者机构：
  
  1.株洲时代电子技术有限公司，湖南　株洲　412001
  2.长沙市轨道交通运营有限公司，湖南　长沙　410000
  3.深圳比一比网络科技有限公司，广东　深圳　518000
- 作者简介：
- 基金信息：
- DOI：10.13890/j.issn.1000-128x.2021.02.016
  CLC：
扫描看全文
Jiangping LUO, Xizhuo YU, Jingwei CAO, et al. Intelligent Rail Flaw Detection System Based on Deep Learning and Support Vector Machine. [J]. Electric Drive for Locomotives (2):100-107(2021)
DOI：

Jiangping LUO, Xizhuo YU, Jingwei CAO, et al. Intelligent Rail Flaw Detection System Based on Deep Learning and Support Vector Machine. [J]. Electric Drive for Locomotives (2):100-107(2021) DOI： 10.13890/j.issn.1000-128x.2021.02.016.

摘要

目前国内钢轨探伤车检测系统都带有自动伤损识别功能，但由于采用了基于既有规则的简单逻辑判断方法，其自动识别的准确率不高，误报较多，伤损漏报的现象时有发生。针对该问题，根据钢轨探伤车所检测数据的特点，提出了基于深度学习与支持向量机的钢轨伤损智能识别系统技术方案；采用深度可分离卷积与选择性搜索相结合的方法进行目标定位；基于人工构建的多维特征，采用支持向量机方式进行伤损图像分类；并通过使用实际线路所测数据中的人工标注样本进行测试，验证了方法的有效性。测试结果表明，该系统在各项技术指标上均表现优异，伤损检出率达到99.8%，误报率降为12%，分类准确率达到95%以上。

Abstract

Currently, detection systems of rail flaw detection vehicles in China have automatic flaw recognition function, which has problems with low accuracy, high false alarm rate and occurrence of underreport because of the adoption of simple logic judgment method based on the existing rules. In view of problems proposed above, according to the characteristics of ultrasonic testing data, an intelligent rail flaw recognition system based on deep learning and support vector machine was proposed in this paper. Depth separable convolution and selective search was adopted to locate the target, and support vector machine method based on manually constructed multidimensional features was used to classify the flaws image in the system. The effectiveness of this measure was verified through the test of manually marked samples from actual running line data. The result showed that the intelligent rail flaw recognition system had excellent performance in many technical indicators of which the flaw detection was 99.8%, the false alarm rate reduced to 12% and the accuracy of classification was over 95%.

关键词

钢轨探伤深度学习支持向量机智能识别轨道交通工程机械

Keywords

rail flaw detectiondeep learningsupport vector machineintelligent recognitionrail transitconstruction machinery

references

石永生. 依靠杂波设置探伤车检测标准的分析探讨[J]. 中国铁路, 2013(5): 67-70.

SHI Yongsheng. Analysis and discussion on test standard of flaw detecting car by clutter setting[J]. Chinese Railways, 2013(5): 67-70.

徐其瑞, 石永生, 许贵阳, 等. GTC-80型钢轨探伤车及其运用[J]. 中国铁路, 2013(11): 55-58.

XU Qirui, SHI Yongsheng, XU Guiyao, et al. GTC-80 flaw detecting vehicles and its application[J]. Chinese Railways, 2013(11): 55-58.

张玉华, 许贵阳, 李培, 等. 钢轨探伤车自主化超声检测系统的关键技术[J]. 中国铁道科学, 2015, 36(5): 131-136.

ZHANG Yuhua, XU Guiyang, LI Pei, et al. Key technology to autonomous ultrasonic detection system on rail flaw detection car[J]. China Railway Science, 2015, 36(5): 131-136.

VIPPARTHY S T, MADHU V N, RAMAKRISHNA G G, et al. Inspection of rails using interface of ultrasonic testing[J]. International Journal of Mechanical Engineering and Robotics Research, 2015, 4(1): 176-184.

黄梦莹, 罗江平, 王文星, 等. 基于图像处理的钢轨伤损分类算法研究[J]. 机车电传动, 2020(4): 41-46.

HUANG Mengying, LUO Jiangping, WANG Wenxing, et al. Research on classification of rail defects based on image processing algorithm[J]. Electric Drive for Locomotives, 2020(4): 41-46.

孙次锁, 刘军, 秦勇, 等. 基于深度学习的钢轨伤损智能识别方法[J]. 中国铁道科学, 2018, 39(5): 51-57.

SUN Cisuo, LIU Jun, QIN Yong, et al. Intelligent detection method for rail flaw based on deep learning[J]. China Railway Science, 2018, 39(5): 51-57.

孙次锁, 张玉华. 基于智能识别与周期检测的钢轨伤损自动预警方法研究[J]. 铁道学报, 2018, 40(11): 140-146.

SUN Cisuo, ZHANG Yuhua. Research on automatic early warning method for rail flaw based on intelligent identification and periodic detection[J]. Journal of the China Railway Society, 2018, 40(11): 140-146.

HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//IEEE Computer Society. 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016(1): 770-778. DOI: 10.1109/CVPR.2016.90http://doi.org/10.1109/CVPR.2016.90.

HOWARD A G, ZHU M L, CHEN B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[J/OL]. (2017-04-17) [2020-12-11]. https://arxiv.org/pdf/1704.04861https://arxiv.org/pdf/1704.04861.

Views

下载量

CSCD

CNKI被引量

Alert me when the article has been cited

提交

Tools

Publicity Resources

No data

Related Author

No data

Related Institution

No data

⁰