Research on Seismic Signal Enhancement Method Based on Deep Learning

At present, human beings are increasingly dependent on oil and gas energy. Due to the influence of a series factors such as exploration technology natural environment, in process actual technology, collected seismic data usually has problem low sampling rate. The lower rate may cause spatial aliasing migration imaging, resulting inaccurate exploration. Therefore, order carry out more accurate exploration, signal enhancement method also puts forward higher requirements. In this paper, will be realized by channel interpolation. A trace interpolation based generative adversarial network is proposed realize it. traditional generally carried complex mathematical transformation or certain assumptions, solve problems existing method. This paper use model deep learning combine with residual network, loss function sum Wasserstein distance content loss. optimized can better traces, frequency been suppressed. signals achieved, theoretical support for provided.