Water supply and drainage pipeline anomaly detection method based on image recognition

By using image recognition technology, a three-channel feature extraction network and neural network model were established, which solved the problems of low detection efficiency and safety risks in water supply and drainage pipelines, and achieved efficient and accurate pipeline anomaly detection and prediction.

CN120932153AInactive Publication Date: 2025-11-11ZIBO KAIHUI WATER SUPPLY EQUIP CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202511041331.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-28
Publication Date
2025-11-11
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing methods for inspecting water supply and drainage pipelines are inefficient and pose safety risks. They are difficult to accurately identify surface texture features and abnormal liquid flow, and cannot determine the location of leaks or other types of anomalies.

Method used

An image recognition-based approach is adopted to acquire video image data through an image sensor and establish a three-channel feature extraction network model, including ResNet-50 for texture feature extraction, 3D convolution for structural deformation feature extraction, and RAFT algorithm for capturing abnormal liquid flow features. Anomaly detection and evaluation are performed by combining bidirectional LSTM neural network and GNN graph neural network.

Benefits of technology

It enables multi-dimensional and comprehensive extraction of pipeline anomalies, improving the accuracy and comprehensiveness of detection, timely detection and prediction of anomaly development trends, shortening detection time, and reducing labor costs and safety risks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120932153A_ABST
    Figure CN120932153A_ABST
Patent Text Reader

Abstract

The invention discloses a water supply and drainage pipeline anomaly detection method based on image recognition. Initial video image data are acquired through an image sensor; establishing a three-channel feature extraction network model, wherein the three-channel feature extraction network model at least comprises a visible light channel, a geometrical shape channel for extracting pipeline structure deformation features through 3D convolution and a dynamic optical flow channel for capturing liquid flow abnormal features based on an RAFT algorithm; inputting an initial video image into the three-channel feature extraction network model to perform feature extraction, establishing an analysis model based on a bidirectional LSTM neural network, inputting feature image data into the analysis model to establish time sequence association, performing abnormal evolution through a GNN graph neural network, performing abnormal type classification according to a pipeline state evaluation index, and obtaining a pipeline state evaluation result. And performing early warning based on a classification result. The detection efficiency is improved, the labor cost and the safety risk are reduced, and operation and maintenance personnel can master the state of the pipeline in time.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • Air source heat pump condensate water discharge abnormity monitoring method based on image enhancement

    CN121582857A

  • Conjunctive action dynamic graph neural network-based sewer fault detection method and system

    CN122527846A