基于YOLOv3的输电工程智能检测与分析技术研究  

Research on intelligent detection and analysis technology of transmission engineering based on YOLOv3

作　　者：周云浩 郭达奇 周鑫 王楠[1] 周迎 ZHOU Yunhao;GUO Daqi;ZHOU Xin;WANG Nan;ZHOU Ying(Electric Power Construction Engineering Consulting Branch,State Grid Beijing Electric Power Company,Beijing 100021,China)

机构地区：[1]国网北京市电力公司电力建设工程咨询分公司,北京100021

出　　处：《电子设计工程》2024年第5期66-69,74,共5页Electronic Design Engineering

基　　金：北京电力公司输变电工程应用项目(SGBJJS00XSJS2100639)。

摘　　要：随着能源供应需求的不断增长以及人工智能技术的持续改进,对输电工程质量智能检测技术的要求也越来越高。针对上述问题,文中开展了基于YOLOv3的输电工程质量智能检测与分析技术研究。通过对采集到的输电工程样本数据进行预处理,经多次迭代计算,删除其中的凸包拐点数据。再将实时提取的单帧图像输入目标识别模型中,统计检测到的目标数量。并对低质量图片进行非线性自适应增强处理,以选取高斯双边函数计算均值信息,进而利用卷积算法完成对输电工程验收图片的滤波处理。在对传统的Faster-RCNN算法加以改进后,将FP-FRCNN模型嵌入密集连接结构中,实现输电工程质量智能检测。算例分析结果表明,所提方法可进行输电工程质量智能检测与分析处理,且检测精度高达99.35%,处理时间则仅为3.21 s。With the continuous growth of energy supply demand and the continuous improvement of artificial intelligence technology,the requirements for intelligent detection technology of transmission project quality are becoming higher and higher.To solve the above problems,this paper carried out the research on intelligent detection and analysis technology of transmission project quality based on YOLOv3,preprocessed the collected transmission project sample data,and deleted the convex inflection point data through multiple iterative calculations.Input the real-time extracted single frame image into the target recognition model,count the number of detected targets,carry out nonlinear adaptive enhancement on the low-quality pictures,select the Gaussian bilateral function to calculate the mean information,and then use the convolution algorithm to filter the acceptance pictures of power transmission projects.After improving the traditional Faster-RCNN algorithm,the FP-FRCNN model is embedded into the dense connection structure to realize the intelligent detection of transmission project quality.The results of example analysis show that the method proposed in this paper can be used for intelligent detection and analysis of transmission project quality.The detection accuracy is as high as 99.35%,and the processing time is as low as 3.21 s.

关 键 词：深度学习 图像处理 目标检测 YOLOv3 智能检测 

分 类 号：TP807[自动化与计算机技术—检测技术与自动化装置] TN99[自动化与计算机技术—控制科学与工程]