The Bozhong 19-2 structure is located in the southwest of Bozhong Depression, with favorable accumulation background. The shallow layer has undergone multiple rounds of exploration, but no large-scale discovery has been made. The main target layer of the lower Ming section is deposited by the meandering river, and the longitudinal and transverse direction of the river changes rapidly, and the contact relationship of the sand body is unusually complex, which presents a weak seismic reflection artifact locally, which does not conform to the objective spreading rule of the sand body and affects the description of the continuous sand body. Through forward analysis, it can be seen that the seismic weak reflection phenomenon is mainly affected by the comprehensive factors such as the absorption and attenuation of the stratum, the thickness of the target layer, the physical properties of the target layer, the velocity of the overlying stratum, the structure of the overlying stratum and the distance between the target layer and the overlying stratum. The amplitude of the original seismic data is weakened and the frequency band is narrowed. In this paper, the compressed sensing optimization reconstruction algorithm is used to reconstruct the original signal, and the reconstructed signal is divided into high frequency body and low frequency body. Then the original signal is used as the input of the intermediate frequency body to realize the attribute fusion, and finally the fusion data body is obtained. The main frequency of the data body is improved, the frequency band is broadened, and the recovery of seismic weak reflection is realized. This method can make the recovery results more in line with the real underground situation without relying on logging data, so as to more objectively reflect the distribution law of shallow sand bodies, which is conducive to the characterization of shallow lithologic traps.