胡志坤,赵超越,王振东,林晨
HU Zhikun,ZHAO Chaoyue,WANG Zhendong,LIN Chen

• 1:智洋创新科技股份有限公司

摘要(Abstract)：

目前输电杆塔的隐患识别通常采用无人机拍摄本体图像,再传输到后台进行分析。这样不利于快速定位隐患位置并及时处理。同时输电杆塔的缺陷检测和识别主要采用基于深度学习的目标检测和分类算法,计算量大,无人机终端的处理器难以达到实时检测的效果,对关键部位小微隐患的检测能力不佳。为了提升无人机在巡检过程中对输电杆塔本体小微隐患检测和分类的准确率和实时性,提出了利用FPN(特征金字塔网络)构建Faster R-CNN(区域卷积神经网络)检测模型的MGFF-KCD(关键部件检测的多粒度特征融合算法)来处理多个粒度的特征信息,提高了算法的准确率。将算法模型在无人机终端进行实时智能分析,选取销钉、绝缘子、防震锤、均压环、鸟巢五类输电杆塔关键部位进行试验,结果表明,该算法在华为Atlas 200芯片设备上可实现每张62 ms的检测速度和88%的准确率。
At present, body images of the transmission towers are taken by UAVs and then transferred to backstage for analysis, which does not favor fast location and timely treatment of hidden trouble. Moreover,object detection and classification algorithm of trouble detection and identification of transmission tower based on deep learning feature large calculation quantity and real-time detection is impossible for the processor at UAV terminal, and there is detection inefficiency of minor hidden trouble in critical parts. To improve the accuracy and real-time performance of the detection and classification of the minor potential hazards in patrol inspection of power transmission by UAV, the paper proposes MGFF-KCD(multi-granularity feature fusion algorithm for key component detection) that constructs a fast R-CNN(area convolutional neural network) detection model based on FPN(feature pyramid network) to process multi-granularity feature information to improve the accuracy of the algorithm. The real-time intelligent analysis of the algorithm model is carried out in the terminal of the UAV. The experiments are carried out in five key parts such as pin, insulator, shockproof hammer, pressure equalizer ring and bird′s nest. The result shows that the algorithm can achieve a detection speed of 62 ms per image and precision of 88% on HUAWEI Atlas 200 chip equipment.

关键词(KeyWords)： 边缘计算;输电杆塔本体隐患;无人机巡检;多粒度融合算法
edge calculation;hidden trouble of transmission tower;UAV patrol;multi-granularity fusion algorithm

Abstract:

Keywords:

基金项目(Foundation): 2019年度山东省重点研发计划（重大科技创新工程）项目（2019JZZY010118）

作者(Author): 胡志坤,赵超越,王振东,林晨
HU Zhikun,ZHAO Chaoyue,WANG Zhendong,LIN Chen

DOI: 10.19585/j.zjdl.202010004

参考文献(References)：

• [1]王万国，田兵，刘越，等.基于RNN的无人机巡检图像电力小部件识别研究[J].地球信息科学学报，2017,19(2):256-263.
• [2]汤踊，韩军，魏文力，等.深度学习在输电线路中部件识别与缺陷检测的研究[J].电子测量技术，2018,41(6):60-65.
• [3]张新影.基于深度学习的输电线均压环检测技术研究[D].北京：北京交通大学，2018.
• [4]陈景文，周鑫，张蓉，等.基于U-net网络的航拍绝缘子检测[J].陕西科技大学学报，2018(4):153-157.
• [5]杨罡，孙昌雯，张娜，等.基于多尺度特征融合的输电线路关键部件检测[J].电测与仪表，2020(3):54-59.
• [6]REN S,HE K,GIRSHICK R,et al.Faster r-cnn:Towards real-time object detection with region proposal networks[C]//Advances in neural information processing systems.2015:91-99.
• [7]EVERINGHAM M,VAN GOOL L,WILLIAMS C K I,et al.The pascal visual object classes(voc)challenge[J].International journal of computer vision,2010,88(2):303-338.
• [8]ZHANG X Y.Accelerating very deep convolutional networks for classification and detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(10):1943-1955.
• [9]孙顶君，赵珊．图像底层特征提取与检索技术[M]．北京：电子工业出版社，2009．
• [10]杨钊．面向图像分类和识别的视觉特征表达与学习的研究[D]．广州：华南理工大学，2014．
• [11]锐松，阿瓜多．特征提取与图像处理[M].2版．李实英，杨高波，译.北京：电子工业出版社，2010.
• [12]Hierarchical neural networks for image interpretation[M].Springer Science&Business Media.2003．
• [13]LIU R T.Image partial blur detection and classification[C]//2008 IEEE Conference on Computer Vision and Pattern Recognition,USA,IEEE,2008:1-8.
• [14]LI D,WANG K D,HU Y H,et al.Detection,Classification,and Tracking of Targets[J].IEEE Signal Processing Magazine,2002,19(2):17-29.
• [15]马广恩，罗火钱，李凤滨，等.无人机航测在文川溪数字流域及智慧水利中的应用[J].浙江水利水电学院学报,2019,31(6):31-34.
• [16]苏超，杨强.基于机器视觉深度学习的绝缘子前端定位方法[J].广东电力，2019,32(9):33-40.
• [17]罗建军，刘振声，龚翔，等.基于无人机图像与迁移学习的线路绝缘子状态评价方法[J].电力工程技术，2019,38(5):30-36.
• [18]周小红，李向欢，石蕾，等.无人机倾斜摄影技术在电力巡线树障检测中的实践应用研究[J].电力大数据，2019,22(80:53-59.
• [19]石书山，谢璐，周筑博，等.基于无人机巡检的输电线路规避系统设计[J].电力大数据，2019,22(8):67-74.
• [20]张仕民，雷泽宇，刘意，等.基于AI的无人机电网智能巡检方案研析[J].四川电力技术，2019,42(3):90-94.
• [21]曹浩楠，王彬，邹国平，等.基于输电线路电磁场分布的无人机测距适应性研究[J].浙江电力，2019,38(3):65-70.
• [22]光明，鄂爱东，钱凤阳，等.基于三维激光扫描技术的输电线路无人机巡检方法分析[J].内蒙古电力技术，2019,37(1):21-24.
• [23]刘志武.轻型无人机在输电带电作业中的应用[J].电力大数据，2018,21(11):83-87.
• [24]仇梓峰，王爽心，李蒙.基于无人机图像的风力发电机叶片缺陷识别[J].发电技术，2018,39(3):277-285.
• [25]钱金菊，张睿卓，王柯，等.输电线路巡检可视化管理系统及其应用[J].广东电力，2018,31(3):109-114.