In this paper seismic inversion was used as a key technique and the seismic wavelet most suitable to the actual underground situation was extracted with the higher-order statistics algorithm. The wavelets extracted in this way and the wavelets extracted with the seismic statistics techniques were used separately for inverting the seismic data of the south part of Tahe oilfield, Tarim basin. The results showed that the resolution of the wavelet inversion with the higher-order statistics method was greatly improved, and the wavelet-inverted section could better distinguish the thin sandstone reservoirs of the upper and lower Carboniferous and their lateral distribution, providing a reliable basis of analysis for the study of thin sandstone reservoirs.