应斌,唐斌,潘俊杰,郭震
YING Bin,TANG Bin,PAN Junjie,GUO Zhen

• 1:国网浙江省电力有限公司台州供电公司
• 2:台州宏达电力建设有限公司

摘要(Abstract)：

电力杆塔是支撑和架空输电线的塔架结构,杆塔检测和无人机智能巡检电网线路相结合已逐渐成为发展趋势。为提升电力设备巡检效率,提出了一种基于YOLOv3改进的目标检测模型,用于无人机巡检输电线路中对于杆塔的实时检测。通过水平镜像、旋转变换、图像错切和颜色变换等方法完成数据扩增,扩大杆塔训练样本。利用K-means聚类算法获取最适合的先验锚框,使其更符合杆塔的形状和比例。改进损失函数,采用GIoU计算边界框回归损失,提升了目标定位的准确性。实验结果表明,相对于其他算法,改进后的YOLOv3模型准确率较高,检测速度达到了每帧65 ms, mAP(多类别平均精度)达到90.8%,可以有效检测到航拍图像的电力杆塔,对无人机巡检输电线路有一定的工程应用价值。
Power towers refer to structures that support transmission lines. Nowadays, a combination of tower detection and UAV intelligent line inspection has gradually become a trend. To improve the inspection efficiency of power equipment, an improved object detection model based on YOLOv3 is proposed for real-time tower detection in transmission lines inspection by UAV. Data amplification is accomplished through multiple methods such as horizontal mirroring, rotation transformation, image cross-cutting, and color transformation to enlarge tower training samplings. K-means clustering algorithm is used to obtain the most suitable anchor box to make it more in line with the shape and proportion of the tower. The loss function is improved, and GIoU is used to calculate the bounding box regression loss, which improves the accuracy of object positioning. The experimental results indicate that the improved YOLOv3 model is of high accuracy with detection speed of 65 ms per frame and the mAP(mean average precision) of 90.8% compared with other algorithms.The algorithm can effectively detect the power tower in aerial images and provides certain engineering application value for UAV inspection of transmission lines.

关键词(KeyWords)： 深度学习;杆塔识别;YOLOv3;目标检测
deep learning;tower identification;YOLOv3;object detection

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司集体企业科技项目（SGTYHT/19-JS-217）

作者(Author): 应斌,唐斌,潘俊杰,郭震
YING Bin,TANG Bin,PAN Junjie,GUO Zhen

DOI: 10.19585/j.zjdl.202105008

参考文献(References)：

• [1]韩冰，尚方.面向无人机输电线路巡检的电力杆塔检测框架模型[J].浙江电力，2016,35(4):6-11.
• [2]柳长安，叶文，吴华，等.融合地理位置信息的电力杆塔检测[J].华中科技大学学报（自然科学版），2013,41（增刊1）:208-210.
• [3]王孝余，李丹丹，张立颖.一种基于监督学习的输电线监测中杆塔的检测方法[J].东北电力技术，2017,38(11):12-14.
• [4]韦汶妍，刘晓立，杨传凯，等.一种基于深度学习网络的输电杆塔智能检测算法[J].广东电力，2020,33(9):166-173.
• [5]GU J,WANG Z,KUEN J,et al.Recent advances in convolutional neural networks[J].Pattern Recognition,2018(77):354-377.
• [6]苏赋，吕沁，罗仁泽.基于深度学习的图像分类研究综述[J].电信科学，2019,35(11):58-74.
• [7]罗元，王薄宇，陈旭.基于深度学习的目标检测技术的研究综述[J].半导体光电，2020,41(1):1-10.
• [8]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.
• [9]GUO G,ZHANG N.A survey on deep learning based face recognition[J].Computer Vision and Image Understanding,2019(189):102805.
• [10]REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:unified,real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:779-788.
• [11]胡志坤，赵超越，王振东，等.基于边缘计算和无人机巡检图像的输电杆塔关键部位隐患智能识别[J].浙江电力，2020,39(10):21-27.
• [12]杨亮，李辉芳，江万寿，等.基于梯度对称性的无人机影像电力线自动提取方法[J].地理信息世界，2017,24(2):94-99.
• [13]邹栋.输电线路中关键部件图像识别及异常检测方法研究[D].保定：华北电力大学，2017.
• [14]李奇.基于深度学习的一阶目标检测算法应用研究[D].长春：吉林大学，2019.
• [15]EVERINGHAM M,VAN L,WILLIAMS C,et al. The pascal visual object classes(voc)challenge[J].International Journal of Computer Vision,2010,88(2):303-338.
• [16]任培铭.基于YOLO的实时目标检测方法研究[D].无锡：江南大学，2019.
• [17]魏宏彬，张端金，杜广明，等.基于改进型YOLOv3的蔬菜识别算法[J].郑州大学学报（工学版），2020(2):7-12.
• [18]ZHOU J,TIAN Y,YUAN C,et al.Improved uav opium poppy detection using an updated yolov3 model[J].Sensors,2019,19(22):4851.
• [19]刘革，郑叶龙，赵美蓉.基于RetinaNet改进的车辆信息检测[J].计算机应用，2020,40(3):854-858.
• [20]REZATOFIGHI H,TSOI N,GWAK J Y,et al.Generalized intersection over union:a metric and a loss for bounding box regression[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2019:658-666.
• [21]付琦.基于深度学习的道路目标检测问题研究[D].北京：北京交通大学，2019.
• [22]陈建促.基于视频的野生动物目标检测算法研究[D].重庆：重庆理工大学，2019.
• [23]高琳，陈念年，范勇.融合多尺度上下文卷积特征的车辆目标检测[J].光电工程，2019,46(4):31-38.
• [24]郭敬东，陈彬，王仁书，等.基于YOLO的无人机电力线路杆塔巡检图像实时检测[J].中国电力，2019,52(7):17-23.
• [25]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.