韩睿,戴哲仁,蒋鹏,李晨,姜雄伟
HAN Rui,DAI Zheren,JIANG Peng,LI Chen,JIANG Xiongwei

• 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

摘要(Abstract)：

现行的集中式缺陷检测系统存在数据量大、检测实时性低等问题，亟需以边缘计算为代表的分布式检测系统。为此，基于单阶段、轻量化的检测模型YOLOv8设计多种模块算法以提升其在电力场景下的检测精度。首先改进Mosaic数据增强算法，引入冲突关系表规避了传统数据增强算法对原始图像数据信息的破坏，增强了图像数据的多样性。然后，使用Res2Net模块代替原有的Bottleneck模块，增强模型对多尺度感知能力的同时也保持了检测模型的轻量化。使用CIoU-NMS算法替代原有的NMS（非极大值抑制）算法，提升了检测模型在聚类去重阶段的召回率与精度。最后，在14个电力场景缺陷上进行实验，均获得了比原有模型更好的检测精度，同时模型在电力场景的缺陷检测速度上亦有提升。
The current centralized defect detection system faces challenges such as large data volume and poor realtime performance, underscoring the pressing need for distributed detection systems, with edge computing as a representative. In response, diverse module algorithms are devised based on the single-stage, lightweight detection model YOLOv8 to boost its accuracy in power scenarios. Firstly, the Mosaic data augmentation algorithm is improved, and a conflict relationship table is introduced to mitigate the damage to original image data information caused by traditional data augmentation algorithms and enhance the diversity of image data. Subsequently, the Res2Net module is employed to replace the original Bottleneck module, reinforcing the model's multiscale perception while retaining its lightweight design. The adoption of the CIoU-NMS algorithm over the existing NMS(non-maximum suppression) algorithm improves the recall rate and precision of the detection model in the clustering and deduplication stage. Finally, experiments on fourteen defects in power scenarios consistently demonstrate the proposed model's superior detection accuracy compared to the original model, accompanied by an accelerated detection speed in defect detection power scenarios.

关键词(KeyWords)： 深度学习;缺陷检测;数据增强;非极大值抑制;YOLOv8
deep learning;defect detection;data augmentation;NMS;YOLOv8

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司科技项目（5211DS220003）

作者(Author): 韩睿,戴哲仁,蒋鹏,李晨,姜雄伟
HAN Rui,DAI Zheren,JIANG Peng,LI Chen,JIANG Xiongwei

DOI: 10.19585/j.zjdl.202404012

参考文献(References)：

• [1]谢庆，张煊宇，王春鑫，等.新一代人工智能技术在输变电设备状态评估中的应用现状及展望[J].高压电器，2022,58(11):1-16.XIE Qing,ZHANG Xuanyu,WANG Chunxin,et al.Application status and prospect of the new generation artificial intelligence technology in the state evaluation of power transmission and transformation equipment[J].High Voltage Apparatus,2022,58(11):1-16.
• [2]周家玉，侯慧娟，盛戈皞，等.状态参量关联规则挖掘及深度学习融合的变压器故障诊断算法[J].高压电器，2023,59(3):108-115.ZHOU Jiayu,HOU Huijuan,SHENG Gehao,et al.Transformer fault diagnosis algorithm based on association rules mining of state parameters and deep learning[J]. High Voltage Apparatus,2023,59(3):108-115.
• [3]马静怡，崔昊杨，张明达，等.基于改进Faster RCNN的小尺度入侵目标识别及定位[J].中国电力，2021,54(3):38-44.MA Jingyi,CUI Haoyang,ZHANG Mingda,et al.Small scale invade-target recognition and location based on improved faster RCNN[J]. Electric Power,2021,54(3):38-44.
• [4] 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.
• [5]张焕坤，李军毅，张斌.基于改进型YOLOv3的绝缘子异物检测方法[J].中国电力，2020,53(2):49-55.ZHANG Huankun,LI Junyi,ZHANG Bin.Foreign object detection on insulators based on improved YOLOv3[J].Electric Power,2020,53(2):49-55.
• [6] REDMON J,ALI F. YOLOv3:An incremental improvement[EB/OL].(2018-03-31)[2019-05-25].https：//pjreddie.com/media/files/papers/YOLOv3.pdf.
• [7]李季，刘乐，牛雨潇，等.融入注意力的YOLOv3绝缘子串识别方法[J].高压电器，2022,58(11):67-74.LI Ji,LIU Le,NIU Yuxiao,et al.YOLOv3 identification method incorporating attention for insulator string[J].High Voltage Apparatus,2022,58(11):67-74.
• [8]李彬，贾滨诚，陈宋宋，等.边缘计算在电力供需领域的应用展望[J].中国电力，2018,51(11):154-162.LI Bin,JIA Bincheng,CHEN Songsong,et al.Prospect of application of edge computing in the field of supply and demand[J].Electric Power,2018,51(11):154-162.
• [9]路艳巧，孙翠英，曹红卫，等.基于边缘计算与深度学习的输电设备异物检测方法[J].中国电力，2020,53(6):27-33.LU Yanqiao,SUN Cuiying,CAO Hongwei,et al.Foreign body detection method for transmission equipment based on edge computing and deep learning[J].Electric Power,2020,53(6):27-33.
• [10]盛戈皞，钱勇，罗林根，等.面向新型电力系统的数字化电力设备关键技术及其发展趋势[J].高电压技术，2023,49(5):1765-1778.SHENG Gehao,QIAN Yong,LUO Lingen,et al. Key technologies and development trend of digital power equipment for new power system[J]. High Voltage Engineering,2023,49(5):1765-1778.
• [11]肖靖，曾锦松，许佳庆，等.基于云边端协同技术的电力安全管控系统设计[J].供用电，2023,40(5):44-52.XIAO Jing,ZENG Jinsong,XU Jiaqing,et al. Design of power safety management and control system based on cloud-edge-device collaboration technology[J]. Distribution&Utilization,2023,40(5):44-52.
• [12]庄莉，刘宝升，王秋琳，等.基于边缘计算的变电站风险预警管控系统设计[J].电子技术应用，2023,49(4):92-97.ZHUANG Li,LIU Baosheng,WANG Qiulin,et al. Design of substation risk early warning and control system based on edge calculation[J]. Application of Electronic Technique,2023,49(4):92-97.
• [13]钱斌，蔡梓文，肖勇，等.基于边缘计算的电表计量系统数据协同检测方案[J].中国电力，2019,52(11):145-152.QIAN Bin,CAI Ziwen,XIAO Yong,et al.Data collaborative detection scheme of electric metering system based on edge computing[J]. Electric Power,2019,52(11):145-152.
• [14]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.HU Zhikun,ZHAO Chaoyue,WANG Zhendong,et al.Intelligent identification of hidden troubles in key parts of transmission towers based on edge computing and UAV inspection images[J]. Zhejiang Electric Power,2020,39(10):21-27.
• [15]高熠，田联房，杜启亮.基于Mask R-CNN的复合绝缘子过热缺陷检测[J].中国电力，2021,54(1):135-141.GAO Yi,TIAN Lianfang,DU Qiliang.Overheating defect detection of composite insulator based on mask R-CNN[J].Electric Power,2021,54(1):135-141.
• [16]刘逸凡，王淑青，庆毅辉，等.基于EfficientDet和双目摄像头的绝缘子缺陷检测[J].中国电力，2021,54(2):156-163.LIU Yifan,WANG Shuqing,QING Yihui,et al.Insulator defect detection based on EfficientDet and binocular camera[J].Electric Power,2021,54(2):156-163.
• [17]应俊，刘迅，曾学仁，等.基于SSD算法优化的风机叶片缺陷检测研究与应用[J].浙江电力，2021,40(8):47-52.YING Jun,LIU Xun,ZENG Xueren,et al.Research and application of turbine blade defect detection based on an optimized SSD algorithm[J]. Zhejiang Electric Power,2021,40(8):47-52.
• [18] NING M,LU Y,HOU W Y,et al.YOLOv4-object:an efficient model and method for object discovery[C]//2021IEEE 45th Annual Computers,Software,and Applications Conference(COMPSAC). July 12-16,2021,Madrid,Spain.IEEE,2021:31-36.
• [19] HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).June 27-30,2016,Las Vegas,NV,USA.IEEE,2016:770-778.
• [20] GAO S H,CHENG M M,ZHAO K,et al. Res2Net:a new multi-scale backbone architecture[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2021,43(2):652-662.
• [21] NEUBECK A,VAN GOOL L. Efficient non-maximum suppression[C]//18th International Conference on Pattern Recognition(ICPR′06). August 20-24,2006,Hong Kong,China.IEEE,2006:850-855.
• [22] ZHENG Z H,WANG P,REN D W,et al.Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J].IEEE Transactions on Cybernetics,2022,52(8):8574-8586.
• [23] HE K M,GKIOXARI G,DOLLáR P,et al. Mask RCNN[C]//2017 IEEE International Conference on Computer Vision(ICCV).October 22-29,2017,Venice,Italy.IEEE,2017:2980-2988.