摘要
为解决供水管道机器人对管道泄漏识别准确率低的问题,根据管道检测机器人工作原理设计了一种改进卷积神经网络(CNN)。将改进CNN与其他方法进行性能对比分析,发现改进CNN网络的召回率和F1值分别为98.23%和97.15%。接着,将改进CNN置于管道检测机器人之中进行识别效果分析,发现基于改进CNN的检测系统对直径为0.2 mm、0.4 mm、0.6 mm和0.8 mm泄漏点的识别准确率分别为93.36%、97.18%、93.27%、94.78%,均优于对比系统。结果表明,改进CNN网络能提高管道泄漏识别的准确性,为供水管道检测相关领域提供一定理论依据。
To address the issue of low accuracy in pipeline leakage identification by water supply pipeline inspection robots,an improved Convolutional Neural Network(CNN)was designed based on the working principles of pipeline inspection robots.A performance comparison analysis was conducted between the improved CNN and other methods,revealing that the recall rate and F1 score of the improved CNN network reached 98.23% and 97.15%,respectively.Subsequently,the improved CNN was integrated into a pipeline inspection robot for an analysis of its identification effectiveness.It was found that the detection system based on the improved CNN achieved identification accuracy rates of 93.36%,97.18%,93.27%,and 94.78% for leakage points with diameters of 0.2 mm,0.4 mm,0.6 mm,and 0.8 mm,respectively,all outperforming the comparative systems.The results indicate that the improved CNN network can enhance the accuracy of pipeline leakage identification,providing a theoretical basis for relevant fields in water supply pipeline inspection.
作者
邹康兵
袁永钦
李启慧
马伟俊
黄启汉
ZOU Kangbing;YUAN Yongqin;LI Qihui;MA Weijun;HUANG Qihan(Guangzhou Water Supply Co.,Ltd.,Guangzhou,Guangdong 510000,China;Fujian Shangrun Precision Instrument Co.,Ltd.,Fuzhou,Fujian 350015,China)
出处
《自动化应用》
2025年第16期57-59,62,共4页
Automation Application
