“In response to the problems of low accuracy in 3D reconstruction of satellite remote sensing stereo images and difficulties in GCPs acquisition under uncontrolled conditions of gully development, the research progress in high-precision terrain modeling was introduced. The research team established a BP neural network method that integrates GEDI laser altimetry data with GF-7 satellite image DSM, explored the relationship between multiple factors and laser measurement high points under uncontrolled conditions, and verified that this method can improve the average elevation accuracy of DSM to 1.63 meters, which is basically equivalent to controlled conditions. It provides a solution to the problem of low terrain modeling accuracy in uncontrolled mining areas and provides new ideas for the promotion and application of domestic high-resolution images and high-precision terrain modeling.”