[1]李 凡,高 瞻,王红斌,等.基于对抗自学习的跨域绝缘子检测算法[J].南京理工大学学报(自然科学版),2020,44(06):651-659.[doi:10.14177/j.cnki.32-1397n.2020.44.06.003]
　Li Fan,Gao Zhan,Wang Hongbin,et al.Cross-domain insulator detection algorithm based onadversarial self-learning[J].Journal of Nanjing University of Science and Technology,2020,44(06):651-659.[doi:10.14177/j.cnki.32-1397n.2020.44.06.003]
点击复制

基于对抗自学习的跨域绝缘子检测算法()
分享到：

《南京理工大学学报》（自然科学版）[ISSN:1005-9830/CN:32-1397/N]

卷:
44卷
期数:
2020年06期
页码:
651-659
栏目:
出版日期:
2020-12-31

文章信息/Info

Title:
Cross-domain insulator detection algorithm based onadversarial self-learning
文章编号:
1005-9830(2020)06-0651-09
作者:
李 凡1; 2; 高 瞻1; 2; 王红斌1; 2; 李 爽1; 2; 庞 健1; 2; 徐开雄1; 2; 余正涛1; 2
昆明理工大学 1.信息工程与自动化学院; 2.云南省人工智能重点实验室,云南 昆明 650500
Author(s):
Li Fan1; 2; Gao Zhan1; 2; Wang Hongbin1; 2; Li Shuang1; 2; Pang Jian1; 2; Xu Kaixiong1; 2; Yu Zhengtao1; 2
1.Faculty of Information Engineering and Automation; 2.Yunnan Key Laboratory of Artificial Intelligence,Kunming University of Science and Technology,Kunming 650500,China
关键词:
对抗学习;  绝缘子;  线路巡检;  图像;  自训练学习;  特征提取器;  玻璃绝缘子;  复合绝缘子
Keywords:
adversarial learning;  insulators;  lines inspection;  images;  self-training learning;  feature extractor;  glass insulators;  composite insulators
分类号:
TP391.1
DOI:
10.14177/j.cnki.32-1397n.2020.44.06.003
摘要:
为实现在海量线路巡检图像中对绝缘子的自动识别,提出一种基于对抗自学习的跨域绝缘子识别方法。该方法由对抗学习和自训练学习2个阶段组成。在对抗学习阶段,通过特征提取器和分类器之间的对抗学习,使模型分别获得对玻璃绝缘子和复合绝缘子具有鲁棒性的分类特征。在自训练学习阶段,首先,采用有标签的玻璃绝缘子样本对模型进行预训练; 然后,将无标签的复合绝缘子样本输入网络,并选择置信度高的样本赋予软标签对模型进行再次训练,使模型最终获得在不同域上的泛化能力。与现有方法相比,该文方法采用分属不同材质的绝缘子样本对深度神经网络进行2个阶段的训练,在有效降低模型训练过程中样本标注量的同时,解决了跨域识别不同材质的绝缘子的问题。
Abstract:
In order to realize the automatic recognition of insulators in inspection images of massive lines,a cross-domain insulator recognition method based on adversarial self-learning is proposed. The method consists of adversarial learning and self-training learning two stages. In the adversarial learning stage,through adversarial learning between the feature extractor and the classifier,the model obtains robust classification features of glass insulators and composite insulators. In the self-training learning stage,firstly,the model is pre-trained with labeled glass insulator samples; then,the unlabeled composite insulator samples are input into the pre-trained model,and the sample with high confidence is selected to give the soft label to retrain the model,so that the model obtains generalization ability in different domains. Compared with the existing methods,the proposed method uses insulator samples of different materials to train the deep neural network in two stages,which effectively reduces the amount of sample annotations in the model training phase and solves the problem of cross-domain identification of different materials.

参考文献/References:

[1] Tudevdagva U,Battseren B,Hardt W,et al. Image processing based insulator fault detection method[EB/OL]https://www. researchgate. net/publication/329508391_Image_Processing_Based_Insulator_Fault_Detection_Method,2020-12-01.
[2]程海燕,韩璞,王迪,等. 一种电网巡检航拍图像中绝缘子定位方法[J]. 系统仿真学报,2017,29(6):1327-1336.
Cheng Haiyan,Han Pu,Wang Di,et al. Location method of insulators in power grid patrol aerial images[J]. Journal of System Simulation,2017,29(6):1327-1336.
[3]姚春羽,金立军,闫书佳. 电网巡检图像中绝缘子的识别[J]. 系统仿真学报,2012,24(9):1818-1822.
Yao Chunyu,Jin Lijun,Yan Shujia. Recognition of insulators in the image of power grid inspection[J]. Journal of System Simulation,2012,24(9):1818-1822.
[4]Redmon J,Divvala S,Girshick R,et al. You only look once:Unified,real-time object detection[EB/OL]. https://arxiv. org/pdf/1506. 02640. pdf,2020-12-01.
[5]Ren Shaoqing,He Kaiming,Girshick R,et al. Faster R-CNN:Towards real-time object detection with region proposal networks[EB/OL]. http://de. arxiv. org/pdf/1506. 01497,2020-12-01.
[6]Cai Zhaowei,Vasconcelos N. Cascade R-CNN:Delving into high quality object detection[EB/OL]. https://arxiv. org/pdf/1712. 00726. pdf,2020-12-01.
[7]Gui Zhongcheng,Li Haifeng. Automated defect detection and visualization for the robotic airport runway inspection[J]. IEEE Access,2020,8:76100-76107.
[8]Aguiar A S,Santos F N D,de Sousa A J M,et al. Visual trunk detection using transfer learning and a deep learning-based coprocessor[J]. IEEE Access,2020,8:77308-77320.
[9]He Yihui,Zhu Chenchen,Wang Jianren,et al. Bounding box regression with uncertainty for accurate object detection[EB/OL]. https://arxiv. org/pdf/1809. 08545. pdf,2020-12-01.
[10]Krizhevsky A,Sutskever I,Hinton G. ImageNet classification with deep convolutional neural networks[EB/OL]. https://www. researchgate. net/publication/267960550_ImageNet_Classification_with_Deep_Convolutional_Neural_Networks,2020-12-01.
[11]LeCun Y,Boser B,Denker J S,et al. Backpropagation applied to handwritten zip code recognition[J]. Neural Computation,1989,1(4):541-551.
[12]Redmon J,Farhadi A. YOLOv3:An incremental improvement[EB/OL]. https://arxiv. org/pdf/1804. 02767. pdf,2020-12-01.
[13]Yang Zhenheng,Mahajan D,Ghadiyaram D,et al. Activity driven weakly supervised object detection[EB/OL]. https://ieeexplore. ieee. org/document/8953417,2020-12-01.
[14]Xie Shaoan,Zheng Zibin,Chen Liang,et al. Learning semantic representations for unsupervised domain adaptation[EB/OL]. http://proceedings. mlr. press/v80/xie18c/xie18c. pdf,2020-12-01.
[15]Caron M,Bojanowski P,Joulin A,et al. Deep clustering for unsupervised learning of visual features[EB/OL]. https://arxiv. org/pdf/1807. 05520. pdf,2020-12-01.
[16]Wang Xudong,Cai Zhaowei,Gao Dashan,et al. Towards universal object detection by domain attention[EB/OL]. https://arxiv. org/pdf/1904. 04402. pdf,2020-12-01.
[17]Ma Xinhong,Zhang Tianzhu,Xu Changsheng. GCAN:Graph convolutional adversarial network for unsupervised domain adaptation[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach,CA,USA:IEEE,2019:8266-8276.
[18]Kim T,Jeong M,Kim S,et al. Diversify and match:A domain adaptive representation learning paradigm for object detection[EB/OL]. https://arxiv. org/pdf/1905. 05396v1. pdf,2020-12-01.
[19]Goodfellow Ian J,Pouget-Abadie J,Mirza M,et al. Generative adversarial networks[EB/OL]. https://arxiv. org/pdf/1406. 2661. pdf,2020-12-01.
[20]陈耀,宋晓宁,於东军. 迭代化代价函数及超参数可变的生成对抗网络[J]. 南京理工大学学报,2019,43(1):35-40.
Chen Yao,Song Xiaoning,Yu Dongjun. Iterative cost function and variable parameter generative adversarial networks[J]. Journal of Nanjing University of Science and Technology,2019,43(1):35-40.
[21]Redmon J,Farhadi A. YOLO9000:Better,faster,stronger[EB/OL]. https://arxiv. org/pdf/1612. 08242. pdf,2020-12-01.
[22]Chen Yuhua,Li Wen,Sakaridis C,et al. Domain adaptive faster R-CNN for object detection in the wild[EB/OL]. https://arxiv. org/pdf/1803. 03243. pdf,2020-12-01.

备注/Memo

备注/Memo:
收稿日期:2020-08-17 修回日期:2020-10-12
基金项目:国家重点研发计划(2018YFC0830105; 2018YFC0830100); 云南电网公司科技项目(YNKJXM20190729)
作者简介:李凡(1986-),男,博士,讲师,硕士生导师,主要研究方向:图像处理及图像识别,E-mail:478263823@qq.com; 通讯作者:王红斌(1983-),男,博士,副教授,硕士生导师,主要研究方向:自然语言处理、机器学习,E-mail:whbin2007@126.com。
引文格式:李凡,高瞻,王红斌,等. 基于对抗自学习的跨域绝缘子检测算法[J]. 南京理工大学学报,2020,44(6):651-659.
投稿网址:http://zrxuebao.njust.edu.cn
更新日期/Last Update: 2020-12-30