Fault-attri-attention: a method for fault identification based on seismic attributes attention

The imaging principle of seismic images is different from natural images, which results in very limited resolution, complex reflection features and strong uncertainty of seismic images. The fault interpretation methods based on seismic attribute analysis have been widely applied in the industry. However, the seismic attribute has inherent limitations and strong multiplicity. In order to overcome the limitations and multiplicity, a method for fault identification based on seismic attributes attention is proposed to enhance the expression ability of seismic multi-attributes fusion in fault identification tasks. Specifically, the fault identification model is proposed to achieve multi-objective joint prediction by fusing seismic multi-attributes. The seismic attributes attention mechanism named Fault-Attri-Attention is proposed to adaptively extract seismic attributes attention according to the difference in contributions of seismic attributes to fault identification tasks, which can obtain the optimal seismic multi-attributes fusion output. The multi-scales TransBlock module is proposed to enhance the feature expression of seismic attributes with different scales. Experimental results show that the fault identification method based on seismic attributes attention can achieve complementary multi-scales features information, which ensures the independence of seismic attributes and the integrity of multivariate information.