海洋开发与管理2025，Vol.42Issue(5)：88-99,12.
基于多模态数据融合的海洋牧场环境实时监测系统设计与实现
Design and Implementation of Marine Ranching Environment Real-Time Monitoring System Based on Multi-Modal Data Fusion
摘要
Abstract
Given the low accuracy of monitoring data and early warning information in the current marine ranch monitoring system,the number and distribution of sensors,outlier correc-tion algorithm and data fusion algorithm were improved and optimized,and a real-time monito-ring system for marine ranch based on multimodal data fusion was designed and implemented.First,the original data monitored by the sensor is sent to the cloud server.The server monitors the abnormal data through the box graph method and uses the algorithm to correct it.Then,the adaptive weighted average algorithm is used to fuse the data monitored by multiple similar sen-sors,and then the locally fused output data is used as the input data of heterogeneous sensor da-ta fusion,and the PSO-RBF algorithm is used for the final fusion.The main innovation of the article lies in using the optimized RBF neural network algorithm to correct the abnormal data.At the same time,the PSO algorithm is introduced into the RBF algorithm model to optimize the deficiencies of the traditional RBF algorithm in the process of heterogeneous data fusion and improve the accuracy of the early warning information of the monitoring system.The experi-mental results show that after correcting the abnormal data,the root mean square error of simi-lar sensor data fusion is reduced by more than 50%;Compared with the traditional RBF algo-rithm,the average absolute error of early warning information is reduced by about 45%;At the same time,the monitoring system runs stably.关键词
数据融合/监测系统/异常值修正/PSO-RBF神经网络/预警信息Key words
Data fusion/Monitoring system/Abnormal value correction/PSO-RBF neural net-work/Warning information分类
海洋学引用本文复制引用
陈小龙,王刚,李明智,王玺华,李文松..基于多模态数据融合的海洋牧场环境实时监测系统设计与实现[J].海洋开发与管理,2025,42(5):88-99,12.基金项目
国家贝类产业技术体系设施养殖岗位(CARS-48) （CARS-48）
辽宁省教育厅科学研究项目(DL202004) （DL202004）
辽宁省自然科学基金计划项目(2023-BSBA-020). （2023-BSBA-020）
