曹志勇,丰佳,毛文利,李治国,张小燕,赖怀景
CAO Zhiyong,FENG Jia,MAO Wenli,LI Zhiguo,ZHANG Xiaoyan,LAI Huaijing

• 1:国网浙江省电力有限公司电力科学研究院
• 2:国网浙江省电力有限公司
• 3:深圳大学计算机与软件学院

摘要(Abstract)：

为了实现对无人机拍摄的电力塔基地面图像的目标检测,首先,建立了一个高分辨率的电力塔基地面图像数据集,并提出了使用基于Faster R-CNN(快速区域卷积神经网络)改进的目标检测算法提高检测准确率。该方法采用ResNet-50作为Faster R-CNN的主干网络代替VGG-16(卷积神经网络结构),在检测过程中加入了多尺度特征融合技术。最后,将改进的Faster R-CNN电力塔基检测算法与YOLOv4(目标检测算法)和SSD(单发多盒探测器)的检测结果进行对比与分析。结果表明,当IoU(重叠度)阈值为0.75时,算法检测结果的平均精度提升了2.9%;当IoU阈值为0.5时,平均精度提升了0.19%。
For object detection of the power tower base images taken by unmanned drones, a high-resolution image dataset of power tower bases was established, and an improved object detection algorithm based on Faster R-CNN(region-based convolutional neural networks) was proposed to improve the detection accuracy.Instead of VGG-16, a convolutional neural network architecture, this method uses ResNet-50 as the backbone network of Faster R-CNN and integrates multi-scale feature fusion techniques into the detection process.Finally, detection results of the power tower base detection algorithm based on this improved Faster R-CNN,YOLOv4, an object detection algorithm, and SSD(single shot multibox detector) were compared and analyzed. The results show that when the IoU(intersection over union) threshold is 0.75, the accuracy of the detection results of this algorithm averages a 2.9 percent increase, and when the IoU threshold is 0.5, the accuracy averages a 0.19 percent increase.

关键词(KeyWords)： 目标检测;快速区域卷积神经网络;电力塔基;通道注意力;多尺度特征融合
object detection;Faster R-CNN;power tower base;channel attention;multi-scale feature fusion

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司科技项目（5211DS18002V）

作者(Author): 曹志勇,丰佳,毛文利,李治国,张小燕,赖怀景
CAO Zhiyong,FENG Jia,MAO Wenli,LI Zhiguo,ZHANG Xiaoyan,LAI Huaijing

DOI: 10.19585/j.zjdl.202111011

参考文献(References)：

• [1]洪倩，陈枫楠，梁冬，等.特高压塔基施工扰动区域快速自动识别研究[J].人民黄河，2019,41(8):99-102.
• [2]李国强，彭炽刚，汪勇，等.基于深度学习的杆塔三维姿态实时估计[J].电子技术应用，2021,47(6):87-91.
• [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]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].计算机工程，2021,47(6):292-298.
• [6]浦东，陈瑞，焦良葆，等.基于轻量级Mask RCNN的均压环歪斜检测[J].电力信息与通信技术，2021,19(5):95-102.
• [7]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.
• [8]应斌，唐斌，潘俊杰，等.基于YOLOv3的电力杆塔检测算法研究[J].浙江电力，2021,40(5):53-59.
• [9]REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:Unified,real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),June 27-30,2016,Las Vegas,USA:779-788.
• [10]LIU W,ANGUELOV D,ERHAN D,et al.SSD:single shot multibox detector[C]//European Conference on Computer Vision,2016:21-37.
• [11]BIAN J,HUI X,ZHAO X,et al.A monocular visionbased perception approach for unmanned aerial vehicle close proximity transmission tower inspection[J].International Journal of Advanced Robotic Systems,2019,16(1):1729881418820227.
• [12]CERóN A,I MONDRAGóN,PRIETO F.Real-time transmission tower detection from video based on a feature descriptor[J].IET Computer Vision,2017,11(1):33-42.
• [13]STEIGER O,LUCAS E,MARET Y.Automatic detection of transmission towers[C]//2014 IEEE SENSORS,November 2-5,2014,Valencia,Spain.IEEE,2014:1034-1037.
• [14]TRAGULNUCH P,CHANVIMALUANG T,KASETKASEM T,et al.High voltage transmission tower detection and tracking in aerial video sequence using object-based im age classification[C]//2018 International Conference on Embedded Systems and Intelligent Technology&International Conference on Information and Communication Technology for Embedded Systems(ICESIT-ICICTES),May 7-9,2018,Khon Kaen,Thailand.IEEE,2018:1-4.
• [15]顾超越，李喆，史晋涛，等.基于改进Faster-RCNN的无人机巡检架空线路销钉缺陷检测[J].高电压技术，2020,46(9):3089-3096.
• [16]黄和锟.基于SSD目标检测算法的多尺度特征融合技术[J].现代信息科技，2020,4(18):122-124.
• [17]CAO G M,XIE X M,YANG W Z,et al.Feature-fused SSD:fast detection for small objects[C]//International Conference on Graphic and Image Processing,2018:381-388.
• [18]GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich feature hierarchies for accurate object detection and se mantic segmentation[C]//2014 IEEE Conference on Com puter Vision and Pattern Recognition,June 23-28,2014,Columbus,USA.IEEE,2014:580-587.
• [19]HE K,ZHANG X,REN S,et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition,June 27-30,2016,Las Vegas,USA.IEEE,2016:770-778.