Catenary pillar number recognition method based on convolutional neural network
-
摘要: 接触网支柱号是铁路供电部门现场运营维护的重要定位参数。基于卷积神经网络的接触网支柱号自动识别方法结合接触网图像的实际特点,对视频图像进行了归一化图像预处理,并对实际的支柱图片进行了卷积神经网络的训练,在支柱号识别确定的过程中考虑了接触网支柱号的分布特点,提高了支柱号识别的准确性。利用实际线路数据进行测试,取得了较好的识别精度和较快的识别速度。通过实验验证,该方法能够辅助铁路基础设施检测系统中缺陷的定位,指导现场运营维修。Abstract: The catenary pillar numberis an important positioning parameter for on-site operation and maintenance of railway power supply department. Based on convolution neural network, the automatic recognition method of catenary pillar number was combined with the actual characteristics of catenary image. The video image was preprocessed by normalization, and the actual pillar image was trained by convolution neural network. The distribution characteristics of catenary pillar number were considered in the process of the identifying. It improves the accuracy of pillar number recognition.The actual railway line data were used to test, and better recognition accuracy and faster recognition speed were obtained. The experimental results show that this method can assist the defect location in the railway infrastructure inspection system and guide on-site operation and maintenance.
-
Key words:
- railway infrastructure inspection /
- catenary /
- pillar number /
- convolution neural network
-
[1] 王登阳,杨超. 高速综合检测列车时空同步技术[J]. 铁路技术创新, 2012(1):16-19. [2] 中华人民共和国铁道部运输局供电部,中国铁道科学研究院, 西南交通大学. 高速铁路供电安全检测监测系统(6C)总体技术规范[S]. 北京:中华人民共和国铁道部, 2012. [3] 吴光龙. 基于机器视觉的接触网巡检关键技术研究[D]. 成都:西南交通大学, 2016. [4] 杨文静,狄岚,梁久祯. 基于多帧杆号字符识别的接触网检测系统[J]. 数据采集与处理, 2016, 31(6):1250-1258. [5] 杜馨瑜,戴鹏,李颖,等. 基于深度学习的铁道塞钉自动检测算法[J]. 中国铁道科学, 2017, 38(3):89-96. [6] 韦璞. 面向视频监控的受电弓与接触网支柱检测[D]. 成都:西南交通大学, 2017. [7] 李建超,张翠兵,柴雪松,等. 基于图像识别技术的隧道衬砌裂缝检测系统研究[J]. 铁道建筑, 2018, 58(1):20-24. [8] 张毅. 深度学习在接触网定位器缺陷检测中的应用[J]. 铁路计算机应用, 2018, 27(3):15-19. [9] 周飞燕,金林鹏,董军. 卷积神经网络综述[J]. 计算机学报, 2017, 40(6):1229-1251. [10] LIU W, ANGUELOV D, ERHAN D, et al.SSD:Single Shot MultiBox Detector[C]//European Conference on Computer Vision 2016.Amsterdam:Springer, 2016:21-37. [11] 雷杰,高鑫,宋杰,等. 深度网络模型压缩综述[J]. 软件学报, 2018, 29(2):251-266. [12] Han Song, Mao Huizi, Dally W.J. Deep compression:compressing deep nueral networks with pruning,trained quantization and Huffman coding[C]//Proc of International Conference on Learning Representations,2016:1-23. [13] Lin Kevin, Yang HuiFang, Hsiao Jenhao,et al.Deep learning of binary hash codes for fast image retrieval large-scale image search query[C]//Proc of IEEE Conference on Computer Vision and Pattern Recogintion.2015:27-35.
点击查看大图
下载:
