基于改进FasterRCNN的绝缘子异常检测Insulator Anomaly Detection Based on an Improved Faster RCNN
何文其,李剑,赵文浩
HE Wenqi,LI Jian,ZHAO Wenhao
摘要(Abstract)：
绝缘子在保证电力传输安全性方面起着重要作用,为提高绝缘子异常检测效率,提出了一种基于改进的FasterRCNN目标检测模型并将其应用于绝缘子检测中。首先,通过无人机采集航拍的图像样本,利用水平翻转、旋转变换、色域变换等通用图像增强方法扩增样本数量,采用Label Img标注工具对各图像样本进行标注,完成绝缘子数据集的搭建。通过K-means聚类计算得到绝缘子数据集的最优先验锚框。针对破碎绝缘子占比小的问题,提出一种拷贝数据增强的方法,解决绝缘子不均衡问题。最后,通过实验对比分析了主干网络、拷贝数据增强和Focal Loss对网络性能的影响,得出结论:所提出的改进FasterRCNN模型提升了绝缘子检测的准确率,通过Resne St101主干网络、拷贝数据增强以及Focal Loss,最终绝缘子数据集的m AP(平均精度均值)达到了68.3%,高于FasterRCNN基线6.8%。
Insulators play an important role in ensuring the safety of power transmission. To improve the efficiency of insulator anomaly detection, an object detection model based on an improved Faster RCNN is proposed and applied to insulator detection. First, the aerial image samples are collected by drones, and the number of samples is amplified by general image augmentation methods such as horizontal flip, rotation transformation and color conversion. Labellmg, a graphical image annotation tool, is used to tagging each image sample to complete the construction of the insulator datasets. The optimal prior anchor frame of insulator datasets is obtained by K-means clustering. In view of the small proportion of broken insulators, this paper proposes a method of copy data augmentation to solve the problem of insulator imbalance. Finally, the influence of backbone network, copy data augmentation and Focal Loss on network performance is compared and analyzed through experiments, which show that the improved Faster RCNN model improves the accuracy of insulator detection. Through the use of Focal Loss, copy data augmentation and ResneSt101 backbone network, the mAP(mean average precision) of the insulator datasets finally reaches 68.3%, 6.8% higher than Faster RCNN baseline.
关键词(KeyWords)：
深度学习;绝缘子识别;FasterRCNN;目标检测
deep learning;insulator identification;Faster RCNN;object detection
基金项目(Foundation): 国网浙江省电力有限公司科技项目(5211TZ1800LL)
作者(Author):
何文其,李剑,赵文浩
HE Wenqi,LI Jian,ZHAO Wenhao
DOI: 10.19585/j.zjdl.202108006
参考文献(References)：
- [1]伋淼,邢镇委,张梦龙.航拍图像中绝缘子缺失故障检测的级联模型[J].科学技术创新,2021(4):36-38.
- [2]申泽浩.航拍架空输电线路绝缘子缺陷检测方法研究[D].天津：天津工业大学,2020.
- [3]易继禹,陈慈发,龚国强.基于改进Faster RCNN的输电线路航拍绝缘子检测[J/OL].计算机工程：1-8[2021-03-17].https://doi.org/10.19678/j.issn.1000-3428.0059872.
- [4]蒋姗,孙渊,严道森.基于深度学习算法的航拍巡检图像的绝缘子识别[J].福州大学学报(自然科学版),2021,49(1):58-64.
- [5]LIU W,ANGUELOV D,ERHAN D,et al.SSD:single shot MultiBox detector[C]//European conference on computer vision.Cham:Springer,2016:21-37.
- [6]BOCHKOVSKIY A,WANG C Y,LIAO H Y MARK.YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2021-04-12].https://arXiv preprint arXiv:2004.10934.
- [7]LAW H,DENG J.CornerNet:detecting objects as paired keypoints[J].International Journal of Computer Vision,2020,128(3):642-656.
- [8]LAW H,DENG J.CornerNet:detecting objects as paired keypoints[J].International Journal of Computer Vision,2020,128(3):642-656.
- [9]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.
- [10]HE K M,GKIOXARI G,DOLLR P,et al.Mask R-CNN[C]//2017 IEEE International Conference on Computer Vision(ICCV).Venice,Italy:IEEE,2017:2980-2988.
- [11]CAI Z W,VASCONCELOS N.Cascade R-CNN:delving into high quality object detection[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City,UT,USA:IEEE,2018:6154-6162.
- [12]LIN T Y,DOLLR P,GIRSHICK R,et al.Feature pyramid networks for object detection[C]//2017 IEEE Confer ence on Computer Vision and Pattern Recognition(CVPR).Honolulu,HI,USA:IEEE,2017:936-944.
- [13]BODLA N,SINGH B,CHELLAPPA R,et al.Soft-NMSimproving object detection with one line of code[C]//2017IEEE International Conference on Computer Vision(ICCV).Venice,Italy:IEEE,2017:5562-5570.
- [14]ZHANG H,WU C R,ZHANG Z Y,et al.ResNeSt:splitattention networks[EB/OL].[2021-04-12].https://arXiv preprint arXiv:2004.08955.
- [15]VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need[EB/OL].[2021-04-12].http://arXiv preprint arXiv:1706.03762,2017.
- [16]CHENG D C,MENG G F,CHENG G L,et al.SeNet:structured edge network for sea-land segmentation[J].IEEE Geoscience and Remote Sensing Letters,2017,14(2):247-251.
- [17]LIN T Y,GOYAL P,GIRSHICK R,et al.Focal loss for dense object detection[C]//2017 IEEE International Conference on Computer Vision(ICCV).Venice,Italy:IEEE,2017:2999-3007.
- [18]ZHANG Z L,SABUNCU M R.Generalized cross entropy loss for training deep neural networks with noisy labels[EB/OL].[2021-04-12].https://arxiv.org/abs/1805.07836v2.
- [19]KRISHNA K,NARASIMHA MURTY M.Genetic Kmeans algorithm[J].IEEE Transactions on Systems,Man,and Cybernetics,Part B(Cybernetics),1999,29(3):433-439.
- [20]王超洋,罗敬一.基于改进FasterRCNN的配网架空线路异常状态检测[J].电力大数据,2020,23(11):9-16.
- [21]王勇,王永旺,郭建勋.基于AI大数据技术的无人机巡线研究[J].电力大数据,2020,23(11):17-23.
- [22]杨瑞娟,岐召阳,连欢.绝缘子挂点单改双横担固定器的研制[J].山西电力,2020(5):61-64.
- [23]郭秀国.高山区域110 kV输变电工程雷击分析及对策[J].山西电力,2020(4):54-58.
- [24]杨旗,曾华荣,黄欢,等.基于BP神经网络的输电线路隐患预放电识别研究[J].电力大数据,2020,23(3):47-54.
- [25]钟黎明,白耀鹏,王鹏宇.直流特高压支柱绝缘子失效原因及检测方法研究[J].山西电力,2019(3):63-68.