人工智能在航检视频缺陷检测分析系统的应用研究  

Research on application of artificial intelligence in aerial inspection video defect detection and analysis system

作　　者：陈晓东 赵敏 邱莉 赵欢 陈泽霖 CHEN Xiaodong;ZHAO Min;QIU Li;ZHAO Huan;CHEN Zelin(Inner Mongolia Dazhi Energy Technology Co.,Ltd.,Hohhot 010000,China)

机构地区：[1]内蒙古达智能源科技有限公司,内蒙古呼和浩特010000

出　　处：《电子设计工程》2023年第14期39-45,共7页Electronic Design Engineering

基　　金：内蒙古科技攻关项目(2019GG304)。

摘　　要：针对航检视频中绝缘子、销钉部件缺陷的问题,采用人工智能检测识别技术,将OCR识别技术、图像显著性特征提取、知识图谱构建、超分辨技术处理以及部件缺陷的深度训练学习与检测等多种技术相融合。通过对图像中杆塔关键部位图像进行检测,构建杆塔关键部件缺陷特征数据库及样本库,实现对杆塔图像中关键部件缺陷的快速、精准识别。经过实践测试,航检图像中绝缘子和销钉缺失缺陷检测平均识别率大于90%,检测速度较人工检测速度提高70%,对输电线路的安全运行有着一定的应用价值,同时为电业局航检视频分析创造经济价值。In view of the defects of insulators and pins in aerial inspection videos,artificial intelligence detection and recognition technology is adopted to integrate OCR recognition technology,image salient feature extraction,knowledge graph construction,super⁃resolution technology processing,and in⁃depth training learning and detection of component defects.Through the detection of the key parts of the image of the tower,the defect feature database and sample database of the key parts of the tower were constructed to realize the rapid and accurate recognition of the defects of the key parts of the tower image.The average recognition rate of insulator and pin defect detection in aerial inspection images is greater than 90%.The detection speed is 70%higher than that of manual detection.It has certain application value for the safe operation of transmission lines and creates economic value for the aerial inspection video analysis of Electric Power Bureau.

关 键 词：OCR识别技术 超分辨技术 知识图谱 对抗网络 深度学习 

分 类 号：TN919.82[电子电信—通信与信息系统]