Application of the transfer learning method in multisource geophysical data fusion

Abstract Using multigeophysical exploration techniques is a common way for deep targets to be explored in complex survey areas. How locate an unknown underground target using multiple datasets great challenge. The useful information the multisource geophysical model can extracted and fused with help of data fusion, which also works well correct interpretation divergence brought on by expert experience, image feature extraction being key step fusion models. Traditionally, this method often used these kinds images, but it significantly reduces efficiency extraction. As result, we propose novel based transfer learning extract features images. First, ResNet50 network initial Owing problems redundancy fuzzy features, Spearman zero phase component analysis achieve reduction enhancement, further improve computational accuracy fusion. Finally, obtained rules that designed current state. algorithm's reliability tested field from Iliamna Volcano. case study demonstrates effectiveness proposed strategy, offers subsurface targets.