基于深度卷积神经网络的铁路接触网鸟窝检测方法研究
Research on Detection of Bird Nests in Overhead Catenary Based on Deep Convolutional Neural Network
- 机车电传动 2019年第4期 页码：126-130
- 作者机构：
1.广西大学 机械工程学院，广西 南宁 530004
2.中车株洲电力机车研究所有限公司，湖南 株洲 412001
- 作者简介：
贺德强（1973—），男，教授，博士生导师，主要从事机车车辆、故障诊断与智能维护、网络化制造等技术的研究工作。
- 基金信息：国家自然科学基金项目(51765006);广西科技攻关项目(桂科攻1598009-6);广西自然科学基金重点项目(2017GXNSFDA198012)
DOI：10.13890/j.issn.1000-128x.2019.04.027
中图分类号：
扫 描 看 全 文
贺德强, 江洲, 陈基永, 等. 基于深度卷积神经网络的铁路接触网鸟窝检测方法研究[J]. 机车电传动, 2019,(4):126-130.
Deqiang HE, Zhou JIANG, Jiyong CHEN, et al. Research on Detection of Bird Nests in Overhead Catenary Based on Deep Convolutional Neural Network[J]. Electric Drive for Locomotives, 2019,(4):126-130.
贺德强, 江洲, 陈基永, 等. 基于深度卷积神经网络的铁路接触网鸟窝检测方法研究[J]. 机车电传动, 2019,(4):126-130. DOI： 10.13890/j.issn.1000-128x.2019.04.027.
Deqiang HE, Zhou JIANG, Jiyong CHEN, et al. Research on Detection of Bird Nests in Overhead Catenary Based on Deep Convolutional Neural Network[J]. Electric Drive for Locomotives, 2019,(4):126-130. DOI： 10.13890/j.issn.1000-128x.2019.04.027.
摘要
鸟类在铁路接触网筑巢一直是造成接触网故障的一个重要原因，目前主要依靠人工巡检的方式确定是否存在鸟窝，不仅工作量大、漏检率高，而且效率低。因此提升接触网鸟窝的检测效率，及时排除隐患，对保障铁路安全运营具有重要的意义。针对此问题，提出了一种基于深度卷积神经网络的Faster R-CNN模型用于接触网鸟窝的自动识别。通过自定义合适的网络结构和参数，经过预训练、RPN网络训练、Fast R-CNN网络训练以及对RPN和Fast R-CNN的联合训练，建立了适合鸟窝检测的Faster R-CNN模型，实现对鸟窝的检测。经试验，Faster R-CNN的准确率为88.5%，每张图片的识别速度为79 ms，通过与传统的HOG方法、DPM方法和卷积神经网络方法进行比较，验证了深度卷积神经网络对铁路接触网鸟窝检测高效性。
Abstract
Nesting on railway catenaries by birds has been an important cause of catenary failure. At present, the inspection for judging the existence of bird’s nests is carried out artificially, which has much shortcomings in heavy working intensity, high missing-inspection rate and low detecting efficiency. So improving detecting efficiency and then removing potential hazards in time is significant to ensure the security operation of railway vehicles. Aiming at such shortcomings, a Faster R-CNN model based on deep convolution neural network was proposed to identify bird’s nests on catenaries automatically. After the steps of customizing appropriate network structure and parameters and going on pre-training, RPN network training, Fast R-CNN network training and the joint training of RPN network and Fast R-CNN network, a Faster R-CNN model which could be implemented bird’s nests detection was established. Experiment has proved that the recognizing accuracy of Faster R-CNN model is 88.5% and the recognizing time of a picture is 79 ms. Compared with the traditional HOG, DPM and convolutional neural network method, it was verify that the Faster R-CNN model based on deep convolutional neural network was efficient on detecting the existence of bird’s nests on railway catenaries.
关键词
深度学习Faster R-CNN接触网鸟窝检测卷积神经网络
Keywords
deep learningFaster R-CNNcatenarynest detectionconvolution neural network
references
曾怡达, 杨金辉, 谭金练, 等. 基于特征匹配的接触网定位支座检测[J]. 机车电传动, 2018 (3): 100-103.
REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
NEUBECK A, VAN GOOL L. Efficient non-maximum suppression[C]//IEEE. 18th International Conference on Pattern Recognition: Volume 3. Hong Kong: IEEE, 2006: 850-855.
ZEILER M D, FERGUS R. Visualizing and understanding convolutional networks[C]// European conference on computer vision. Computer Vision – ECCV 2014. Belin: Springer, 2014: 818-833.
SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J/OL]. Computer Science . arXiv:1409.1556, 2014 [2018-09-01]. https://arxiv.org/abs/1409.1556https://arxiv.org/abs/1409.1556.
RUSSAKOVSKY O, DENG J, SU H, et al. Imagenet large scale visual recognition challenge[J]. International Journal of Computer Vision, 2015, 15(3): 211-252.
RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986 (323): 533-536.
BOTTOU L. Stochastic gradient descent tricks[C]//MONTAVON G, ORR G B, MÜLLERr KR. Neural networks: Tricks of the trade. Berlin: Springer, 2012: 421-436.
GIRSHICK R. Fast R-CNN[J/OL]. Computer Science, arXiv:1504.08083 [2018-09-01]. https://arxiv.org/abs/1504.08083https://arxiv.org/abs/1504.08083. DOI: 10.1109/ICCV.2015.169http://doi.org/10.1109/ICCV.2015.169.
UIJLINGS J R R, SANDE K E A, GEVERS T, et al. Selective search for object recognition[J]. International Journal of computer vision, 2013, 104(2): 154-171.
DALAL N, TRIGGS B. Histograms of oriented gradients for human detection[C]//IEEE. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition: Volume 1. San Diego: IEEE, 2005: 886-893.
FELZENSZWALB P, MALLESTER D, RAMANAN D. A discriminatively trained, multiscale, deformable part model[C/OL]//IEEE. 2008 IEEE Conference on Computer Vision and Pattern Recognition. Alaska: IEEE, 2008[2018-09-01]. https://ieeexplore.ieee.org/document/4587597https://ieeexplore.ieee.org/document/4587597. DOI: 10.1109/CVPR.2008.4587597http://doi.org/10.1109/CVPR.2008.4587597.
0
浏览量
4
下载量
0
CSCD
9
CNKI被引量
- 网络PDF下载
- Meta-XML下载
- BibTeX下载
- Endnote下载
- RefMan 下载
- NoteExpress下载
- NoteFirst下载
关联资源
相关文章
相关作者
相关机构
- 技术支持由北京北大方正电子有限公司提供 湘ICP备20010475号-1
京公网安备11010802024621 - 本系统建议在Chrome、 IE9+ 以上版本浏览器阅读本站内容，360浏览器请切换至极速模式
- Cookies帮助我们提供服务并提供个性化体验。使用本网站，即表示您同意我们使用Cookies