Abstract: The complexity of the stress field of deep geologic body can be attributed to the nonlinear characteristics of the rock mass in highland and high karst water pressure overburden environments, as well as the reconstruction of local tectonic stress field. A nonlinear inversion and optimization method that can analyze the characteristics of the tectonic information at deep stress measurement points is established considering the mechanical response mechanism of multi-field coupling for rock bodies. The coupling effect of temperature-percolation-ground stress field in deep geological body and the mechanism of temperature & percolation fields affecting the ground stress field are systematically analyzed. And the triaxial compression test of the deep rock body in Sanshan Island gold mining area under different surrounding pressures and pore water pressures is carried out to explore the non-linear variation law of physical and mechanical parameters in deep rock body with different pore water pressure and surrounding pressure. The nonlinear feature extended iteration algorithm (NCEI algorithm) is proposed to optimize the overall fitting of the feature information for weakly nonlinear and continuous parts of the measured in-situ stress in each region, and to enhance the resolution of the feature information for strongly nonlinear and discontinuous parts of the in-situ stress. Based on the secondary development interface of ABAQUS software, a muti-coupling calculation subroutine is written to realize the coupled iterative calculation of in-situ stress field-temperature field-percolation field. Then, the deep in-situ stress field in Sanshan Island mining area is inversely calculated and analyzed. The inversion results show that the average inversion accuracy of the in-situ stress components at each measurement point using the NCEI algorithm and considering the multi-field coupling effect of rock mass is 85.62%, which is higher than that of the BP neural network algorithm and the NCEI algorithm without considering the coupling effect of rock mass. Therefore, the algorithm proposed in the paper can fully consider the characteristics of the deep geological environment, and provide a new idea for the inversion and construction of deep complex in-situ stress field.