Research On Noise Interference Suppression Of Desert Seismic Data Based On CAGAN Network

Energy is vital to the development of the country.Our country's energy demand will inevitably continue to grow.In this context,it is particularly important to explore the unexplored energy and mineral resources in our country.There are still many unexplored areas in our country's Tarim Basin,which are likely to contain a large amount of unexplored energy and mineral resources.Therefore,it is necessary to explore this area.Due to the unique topography and climate of the Tarim Basin,the effective signal energy in desert seismic data obtained by seismic exploration technology is extremely weak,and the energy of noise interference is extremely strong,resulting in a very low signal-to-noise ratio of seismic data.At the same time,since the effective signals and noise interference in desert seismic data are distributed in the low frequency band,the phenomenon of spectral aliasing causes many effective signals to be covered by noise interference.Therefore,the problems of effectively suppressing noise interference,obtaining high-quality effective event signals,and improving the signal-to-noise ratio of seismic data are a major difficulty in the field of seismic signal processing.Because the noise interference in desert seismic data has complex characteristics such as high amplitude,low frequency,nonlinearity and non-stationarity,although the traditional noise interference suppression method can suppress the noise interference in desert seismic data to a certain extent,it is not effective for the effective signal.The extraction effect is not ideal,and the "three highs" requirements of "high signal-to-noise ratio","high resolution" and "high fidelity" cannot be achieved.In recent years,related methods of deep learning have emerged in the field of signal processing.Therefore,to effectively suppress the noise interference in desert seismic data and obtain high-quality seismic data,this paper proposes a CAGAN network-based method for the above problems.A new noise suppression method for desert seismic data.The CAGAN network is a new deep learning network constructed by introducing the Class Activation Map(CAM)applied in classification and detection into the Generative Adversarial Network(GAN).CAM can obtain attention weights in sample data through convolutional layers and Global Average Pooling(GAP).The introduction of these attention weights into the network can improve the overall feature extraction ability and optimize the output results.The CAGAN network combines the generator network with the CAM to analyze and train a large amount of sample data,obtain the common features of the deep data,and perform multiple parameter optimization and iteration through the feedback of the loss function and the discriminator network.A denoising model for effectively suppressing noise interference in desert seismic data.Using the trained network model to denoise desert seismic data can effectively suppress noise interference and obtain high-quality seismic data.In the experimental part of this paper,the method of this paper is applied to the processing of synthetic desert seismic data and field desert seismic data.The experimental results show that the method of this paper can well suppress the noise interference in desert seismic data and obtain high-quality seismic data.The method is compared with the traditional noise suppression algorithm and the deep learning denoising algorithm.The comparative experimental results show that the proposed method can suppress the noise interference in desert seismic data more effectively than the comparative method,and the effective signal amplitude preservation and distinguishability in the processing results are also better.