海洋测绘2024，Vol.44Issue(1)：16-20,5.DOI:10.3969/j.issn.1671-3044.2024.01.004
基于地形单元分区的BP神经网络反演海底地形
Inversion of seafloor topography using BP neural network based on geographical unit partition
摘要
Abstract
Correlation exists between seafloor topography and the ocean gravity field in a certain frequency domain.Traditional seafloor topography inversion methods mainly consider the linear relationship between seafloor topography and gravity data,while ignoring the influence of nonlinear terms on the inversion results.Based on this,this paper proposes a seafloor topography inversion method based on subdivision of seafloor units and BP neural network,which improves the accuracy of seafloor topography inversion.The area near the East Pacific Rise at 90° E in the northern East Indian Ocean(83°E-92°E,10°N-10°S)is selected as the experimental area in this paper.By integrating different data such as bathymetry and satellite altimeter gravity,according to the topographic characteristics,the experiment was divided into sea basin area and ridge area,and the BP neural network method was used to perform seabed topography inversion respectively,the seafloor topography of the subregion is fused to further construct a 1'× 1'local seafloor topography model for the experimental area.The accuracy of the model is evaluated through actual measurement data.The results show that the model average relative error accuracy based on the subdivision of seafloor units in this paper can reach 1.45％,and the root mean square error of the inversion results is reduced by 42 m compared with the un-subdivided inversion results,which verifies the effectiveness and feasibility of the proposed method.关键词
海底地形/BP神经网络/分区反演/重力异常/垂直重力梯度异常Key words
seafloor topography/BP neural network/partition inversion/gravity anomaly/vertical gradient of gravity anomaly分类
天文与地球科学引用本文复制引用
范美琦,陈义兰,孙贺元,付延光,周兴华..基于地形单元分区的BP神经网络反演海底地形[J].海洋测绘,2024,44(1):16-20,5.基金项目
国家自然科学基金(42104035) （42104035）
中央级公益性科研院所基本科研业务费专项资金(2023Q05) （2023Q05）
山东省自然科学基金(ZR2020QD087) （ZR2020QD087）
自然资源部海洋测绘重点实验室开放研究基金资助课题(2021B01). （2021B01）
