Abstract: In order to solve the problem of poor imaging effect of near surface karst cave and goaf location seismic data in seismic exploration, this paper summarizes the frequency, energy and signal-to-noise ratio characteristics of seismic data in this area using forward modeling combined with actual data analysis. On this basis, a set of corresponding processing methods are proposed, including micro log constrained tomography static correction, weak signal extraction and compensation, pre-stack five dimensional data regularization and structure constrained grid tomography velocity modeling. Through the application in Nanchuan area, the signal-to-noise ratio and wave group continuity of seismic data are obviously improved. It is proved that this method can effectively improve the quality of seismic data of karst caves and goafs in seismic exploration.