DeepLabv3+ Network-based Infrared Image Segmentation Method for Current Transformer
-
摘要: 红外图像智能分析是变电设备故障诊断的一种有效方法,目标设备分割是其关键技术。本文针对复杂背景下电流互感器整体分割难的问题,采用基于ResNet50的DeepLabv3+神经网络,用电流互感器的红外图像训练语义分割模型的方法,对收集到的样本采用限制对比度自适应直方图均衡化方法实现图像轮廓增强,构建样本数据集,并运用图像变换扩充样本数据集,搭建语义分割网络训练语义分割模型,实现电流互感器像素与背景像素的二分类。通过文中方法对420张电流互感器红外图像测试,结果表明,该方法的平均交并比(Mean Intersection over Union, MIoU)为87.5%,能够从测试图像中精确分割出电流互感器设备,为后续电流互感器的故障智能诊断做铺垫。
-
关键词:
- 红外图像 /
- 电流互感器 /
- ResNet50 /
- DeepLabv3+ /
- 语义分割
Abstract: Infrared image intelligent analysis is an effective method for the fault diagnosis of transformer equipment, and its key technology is target device segmentation. In this study, aiming to address the difficulty in overall segmentation of current transformers with complex backgrounds, the DeepLabv3+ neural network based on ResNet50 was applied to train the semantic segmentation model with infrared image of CT. The collected samples were enhanced by the limited contrast adaptive histogram equalization method, and a sample dataset was constructed. The sample dataset was expanded by image distortion, and a semantic segmentation network was built to train the semantic segmentation model to realize the binary classification of current transformer pixels and background pixels. The test results of 420 current transformer infrared images showed that the MIOU of this method is 87.5%, which can accurately divide the current transformer equipment from the test images and lay a foundation for the subsequent intelligent fault diagnosis of current transformers.-
Key words:
- infrared image /
- current transformer /
- ResNet50 /
- DeepLabv3+ /
- semantic segmentation
-
表 1 多种模型测试数据表
Table 1. Test data table of various models
Model Categories Accuracy IoU MIoU DeepLabv3+(ResNet50) CT 0.86 0.77 0.855 Background 0.95 0.94 DeepLabv3+(ResNet18) CT 0.81 0.72 0.81 Background 0.92 0.90 SegNet CT 0.67 0.44 0.615 Background 0.86 0.79 FCN-8s CT 0.75 0.63 0.74 Background 0.89 0.85 表 2 基于ResNet50的DeepLabv3+模型加入后处理前后测试对比
Table 2. Comparison of tests before and after the addition of the DeepLabv3+ model based on ResNet50
Model Categories Accuracy IoU MIoU DeepLabv3+(ResNet50) CT 0.86 0.77 0.855 Background 0.95 0.94 Our algorithm CT 0.87 0.79 0.875 Background 0.97 0.96 -
[1] 王小芳, 毛华敏. 一种复杂背景下的电力设备红外图像分割方法[J]. 红外技术, 2019, 41(12): 1111-1116. http://hwjs.nvir.cn/article/id/hwjs201912004WANG Xiaofang, MAO Huamin. Infrared Image Segmentation Method for Power Equipment in Complex Background[J]. Infrared Technology, 2019, 41(12): 1111-1116. http://hwjs.nvir.cn/article/id/hwjs201912004 [2] GONG X, YAO Q, WANG M, et al. A deep learning approach for oriented electrical equipment detection in thermal images[J]. IEEE Access, 2018: 1-1. Doi: 10.1109/ACCESS.2018.2859048. [3] 康龙. 基于红外图像处理的变电站设备故障诊断[D]. 北京: 华北电力大学, 2016.KANG Long. Substation equipment fault diagnosis based on infrared image processing[D]. Beijing: North China Electric Power University, 2016. [4] 曾亮. 基于红外图像的变电站设备故障精准定位方法的研究[D]. 重庆: 重庆理工大学, 2019.ZENG Liang. Research on precise fault location method of substation equipment based on infrared image[D]. Chongqing: Chongqing University of Technology, 2019. [5] ZOU H, HUANG F. A novel intelligent fault diagnosis method for electrical equipment using infrared thermography[J]. Infrared Physics & Technology, 2015, 73: 29-35. http://www.onacademic.com/detail/journal_1000038244612810_35a3.html [6] 王旭红, 李浩, 樊绍胜, 等. 基于改进SSD的电力设备红外图像异常自动检测方法[J]. 电工技术学报, 2020, 35(S1): 302-310. https://www.cnki.com.cn/Article/CJFDTOTAL-DGJS2020S1034.htmWANG Xuhong, LI Hao, FAN Shaosheng, et al. Infrared image anomaly automatic detection method for power equipment based on improved single shot multi box detection[J]. Transactions of China Electrotechnical Society, 2020, 35(S1): 302-310. https://www.cnki.com.cn/Article/CJFDTOTAL-DGJS2020S1034.htm [7] 林颖, 郭志红, 陈玉峰. 基于卷积递归网络的电流互感器红外故障图像诊断[J]. 电力系统保护与控制, 2015, 43(16): 87-94. https://www.cnki.com.cn/Article/CJFDTOTAL-JDQW201516013.htmLIN Ying, GUO Zhihong, CHEN Yufeng. Convolutional-recursive network based current transformer infrared fault image diagnosis[J]. Power System Protection and Control, 2015, 43(16): 87-94. https://www.cnki.com.cn/Article/CJFDTOTAL-JDQW201516013.htm [8] 刘云鹏, 裴少通, 武建华, 等. 基于深度学习的输变电设备异常发热点红外图片目标检测方法[J]. 南方电网技术, 2019, 13(2): 27-33. https://www.cnki.com.cn/Article/CJFDTOTAL-NFDW201902006.htmLIU Yunpeng, PEI Shaotong, WU Jianhua, et al. Deep learning based target detection method for abnormal hot spots infraredimages of transmission and transformation equipment[J]. Southern Power System Technology, 2019, 13(2): 27-33. https://www.cnki.com.cn/Article/CJFDTOTAL-NFDW201902006.htm [9] 王晨. 基于深度学习的红外图像语义分割技术研究[D]. 上海: 中国科学院大学(中国科学院上海技术物理研究所), 2017.WANG Chen. Research on infrared image semantic segmentation technology based on deep learning[D]. Shanghai: University of Chinese Academy of Sciences (Shanghai Institute of Technical Physics, Chinese Academy of Sciences), 2017. [10] 邝辉宇, 吴俊君. 基于深度学习的图像语义分割技术研究综述[J]. 计算机工程与应用, 2019, 55(19): 12-21, 42. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201919003.htmKUANG Huiyu, WU Junjun. Survey of image semantic segmentation based on deep learning[J]. Computer Engineering and Applications, 2019, 55(19): 12-21, 42. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201919003.htm [11] LONG J, Shelhamer E, Darrell T. Fully convolutional net-works for semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015: 3431-3440. [12] 袁铭阳, 黄宏博, 周长胜. 全监督学习的图像语义分割方法研究进展[J]. 计算机工程与应用, 2021, 57(4): 43-54. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG202104007.htmYUAN Mingyang, HUANG Hongbo, ZHOU Changsheng. Research progress of image semantic segmentation based on fully supervised learning[J]. Computer Engineering and Applications, 2021, 57(4): 43-54. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG202104007.htm [13] Badrinarayanan V, Kendall A, Cipolla R. SegNet: a deep convolutional encoder-decoder architecture for image segmentation[C]//IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481-2495, DOI: 10.1109/TPAMI.2016.2644615. [14] Garcia-Garcia A, Orts-Escolano S, Oprea S, et al. A review on deep learning techniques applied to semantic segmentation[J/OL]. Computer Vision and Pattern Recognition, 2017. https://arxiv.org/abs/1704.06857. [15] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-ResNet and the impact of residual connections on learning[J/OL]. Computer Vision and Pattern Recognition, 2016. https://arxiv.org/abs/1602.07261. [16] CHEN L C, Papandreou G, Kokkinos I, et al. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848. doi: 10.1109/TPAMI.2017.2699184 [17] 刘致驿, 孙韶媛, 任正云, 等. 基于改进DeepLabv3+的无人车夜间红外图像语义分割[J]. 应用光学, 2020, 41(1): 180-185. https://www.cnki.com.cn/Article/CJFDTOTAL-YYGX202001031.htmLIU Zhiyi, SUN Shaoyuan, REN Zhengyun, et al. Semantic segmentation of nocturnal infrared images of unmannedvehicles based on improved DeepLabv3+[J]. Journal of Applied Optics, 2020, 41(1): 180-185. https://www.cnki.com.cn/Article/CJFDTOTAL-YYGX202001031.htm [18] 于天河, 赵树梅, 兰朝凤. 结合视觉特性的红外图像增强方法[J]. 激光与红外, 2020, 50(1): 124-128. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW202001024.htmYU Tianhe, ZHAO Shumei, LAN Chaofeng. Infrared image enhancement method combining visual characteristics[J]. Laser & Infrared, 2020, 50(1): 124-128. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW202001024.htm [19] Zuiderveld Karel. Contrast Limited Adaptive Histograph Equalization[J]. Graphic Gems IV. San Diego: Academic Press Professional, 1994: 474-485. DOI: 10.1016/B978-0-12-336156-1.50061-6. [20] WONG S C, Gatt A, Stamatescu V, et al. Understanding data augmentation for classification: when to warp?[C/OL]//International Conference on Digital Image Computing: Techniques and Applications (DICTA). IEEE, 2016. https://arxiv.org/pdf/1609.08764.pdf. [21] Csurka G, Larlus D, Perronnin F. What is a good evaluation measure for semantic segmentation?[C/OL]//BMVC, 2013. http://www.bmva.org/bmvc/2013/Papers/paper0032/abstract0032.pdf.