基于魏格纳分布的智能供水管道泄漏分类方法  

Intelligent classification method of water supply pipeline leak based on Wigner-Ville distribution

作　　者：张晟豪 黄建华 郭闯 钟华[1] ZHANG Shenghao;HUANG Jianhua;GUO Chuang;ZHONG Hua(School of Communication Engineering,Hangzhou Danzi University,Hangzhou Zhejiang 310018,China)

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2022年第6期27-32,共6页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：国家自然科学基金资助项目(61301248);浙江省公益技术研究计划资助项目(LGG18F010009);浙江省教育厅一般科研资助项目(Y201942113)。

摘　　要：提出一种基于魏格纳分布的智能供水管道泄漏分类方法。首先,采用魏格纳分布提取混合泄漏信号的时频特征,形成时频分布二维图;然后,进行灰度化预处理,并采用中值滤波来消除混合泄漏信号中的噪声干扰;最后,对经过滤波提取的泄漏信号进行恒虚警处理,减小了时频图像的特征维度,实现了小样本、轻量级的卷积神经网络(Convolutional Neural Network, CNN)训练。实验结果表明,提出方法的泄漏分类准确率达到95.0%以上。An intelligent method of pipeline leak classification based on Wigner-Ville distribution is proposed. Firstly, Wigner-Ville distribution is used to extract the time-frequency characteristics of the mixed leakage signal, and a two-dimensional time-frequency distribution graph is formed. Then, the noise and interference in the mixed leakage signal is eliminated by gray pretreatment and median filtering. Finally, constant false alarm rate processing is performed on the extracted leakage signal features to reduce the feature dimension of the time-frequency image and achieve small sample and lightweight Convolutional Neural Network(CNN) training. The experimental results show that the leak classification accuracy of the proposed method is over 95.0%.

关 键 词：魏格纳分布 卷积神经网络 泄漏检测 分类识别 

分 类 号：TU991.61[建筑科学—市政工程]