A method for detecting concentration of ocean suspended matter by combining all-weather remote sensing and underwater sound

By employing a federated learning architecture combining all-weather remote sensing and underwater acoustic sensor networks, and integrating the CNN-LSTM-Attention model, the problem of insufficient generalization ability in suspended matter concentration detection models was solved. This enabled efficient detection of deep water bodies and optimized data transmission, thereby improving detection accuracy and system efficiency.

CN122409443APending Publication Date: 2026-07-17YANGJIANG BRANCH OF CHINA UNITED NETWORK COMMUNICATIONS CO LTD +2
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
YANGJIANG BRANCH OF CHINA UNITED NETWORK COMMUNICATIONS CO LTD
Filing Date
2026-03-04
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing suspended solids concentration detection models lack generalization ability, making it difficult to cover deep water bodies. Sensor networks also face high data transmission pressure, hindering efficient real-time modeling.

Method used

A federated learning architecture combining all-weather remote sensing and underwater acoustic sensor networks is adopted. The global model is initialized through a central server, and the underwater acoustic sensor nodes perform local training and parameter uploading. The global model is then aggregated and updated. The CNN-LSTM-Attention model is used to extract underwater acoustic signal features and introduce remote sensing inversion labels to achieve distributed data collaborative modeling.

Benefits of technology

It improves the coverage and accuracy of detection, reduces data transmission latency and server load, and enhances the model's generalization ability and robustness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122409443A_ABST
    Figure CN122409443A_ABST
Patent Text Reader

Abstract

The application discloses a kind of ocean suspended matter concentration underwater acoustic federal detection method combined with all-weather remote sensing, belong to marine environmental monitoring technical field.The method includes: center server initializes CNN-LSTM-Attention global model and is issued to each underwater acoustic sensing node;Node utilizes local collection underwater acoustic signal and temperature, salinity, depth data, combines concentration label generated by remote sensing image inversion and matching, trains local model;Node uploads after model parameter compression training;Server carries out effectiveness evaluation and aggregation to parameter, updates global model and is issued again, carries out multiple rounds of federal iteration optimization;Finally, the trained global model is distributed to all nodes for online concentration prediction.The application fuses multi-source data and federal learning, solves the problem of insufficient generalization ability, large data transmission and calculation pressure of traditional method, realizes high-precision, high-efficiency, large-scale distributed collaborative detection of ocean suspended matter concentration.
Need to check novelty before this filing date? Find Prior Art