基于图像识别的电力电缆隧道结构病害检测  

Power Cable Tunnel Structure Disease Detection Based on Image Recognition

作　　者：徐欣 苏梦婷 陈彦 操卫康 张军 XU Xin;SU Meng-ting;CHEN Yan;CAO Wei-kang;ZHANG Jun(Suzhou Power Supply Company State Grid Jiangsu Electric Power Company,Suzhou 215000 China;Guodian Nari Nanjing Control System Co.,Ltd.,Nanjing 210000 China)

机构地区：[1]国网江苏省电力有限公司苏州供电分公司,江苏苏州215000 [2]国电南瑞南京控制系统有限公司,江苏南京210000

出　　处：《自动化技术与应用》2022年第11期23-26,共4页Techniques of Automation and Applications

基　　金：国网江苏省电力有限公司科技项目(J2020048)。

摘　　要：电力电缆隧道结构跨度较长,实际量化形成的灾害指标精度较差,导致灾害检测时得到的裂缝面积数值较大,针对该问题,提出基于图像识别的电力电缆隧道结构病害检测方法。采用Imagewarel3获取隧道结构图像,获取数值化的隧道空间,引用模糊算子迭代骤变函数,控制量化的指标精度,量化电力电缆隧道病害指标,实现对电力电缆隧道结构病害的检测。随机选定一处存在病害的电力电缆隧道结构,用基于深度学习的检测方法、基于条件随机场的检测方法及所设计的病害检测方法进行实验,结果表明所设计方法检测到的裂缝面积数值最小,且符合隧道的周期演变规律。The structure span of power cable tunnels is long,and the accuracy of disaster index formed by actual quantification is poor.As a result,the crack area value obtained during disaster detection is large.To solve this problem,an image-based detection method for structural disease of power cable tunnels is designed.Imagewarel3 is used to obtain the tunnel structure image,obtain the numeri-cal tunnel space,use the fuzzy operator to iterate the step function,control the quantified index accuracy,quantify the index of power cable tunnel disease,and realize the detection of power cable tunnel structure disease.A power cable tunnel structure with disease is randomly selected.The experiment is carried out with the method of depth learning based detection,conditional ran-dom field based detection and disease detection.The result shows that the crack area detected by the method is the smallest and conforms to the regularity of tunnel cycle evolution.

关 键 词：图像识别 病害检测 指标精度 裂缝面积 周期演变规律 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]