Abstract: Channel sandstones exhibit distinct azimuthal anisotropy, enabling anisotropic inversion to more accurately characterize and predict channel sand bodies. However, direct inversion frequently yields unstable results due to an excessive number of parameters to be inverted. In this study, the anisotropic characteristics of channel sand bodies were first determined based on logging data. Comparison of various amplitude variation with offset (AVO) approximate equations demonstrates that the three-term Rűger equation outperformed others. To address the instability of anisotropic inversion, this study proposed a two-step inversion strategy based on wide-azimuth, high-density seismic data. Specifically, the gradient terms of various azimuth sectors were first derived through sector-specific inversion. The relative values of elastic parameters were then calculated using robust linear inversion algorithms. Finally, the absolute values of elastic parameters were determined through machine learning-based high-resolution fusion. The distribution of thin channel sandstone reservoirs predicted by combining petrophysical analysis agreed well with log data and geological understanding. The results of this study can serve as a guide for the prediction and anisotropic inversion of thin channel sandstone reservoirs in similar areas.