Soil moisture capacitive sensors exhibited issues with low detection accuracy and limited applicability of calibration models to saline soils. In order to solve these problems, a calibration method for moisture capacitive sensors based on conductivity compensation of the relative permittivity was proposed. Firstly, a logarithmic model between soil relative permittivity and sensor output was established in the standard solution, and the R2 of the model was 0.983. Furthermore, a regression calibration model of dielectric constant compensation for standard solution conductivity was established through the binary quadratic regression analysis, and the R2 of the model was 0979. The calibration model was about relative permittirity, the output voltage and conductivity. Secondly, the thirdorder polynomial calibration model of the relationship between soil volume water content and relative permittivity was established according to the special calibration of soil sensor in soil sample, and the R2 of the model was 0.996. Finally, the above two-step calibration equation was verified by soil measurement. When the soil conductivity ranged between 0dS/m and 2dS/m, the detection error range of soil volumetric moisture content was reduced from 0.0383m3/m3 to 0.0127m3/m3, and the maximum relative error was reduced from 12.0200% to 6.2241%. The results indicated that, in similar soils with different conductivities, using the calibration method for moisture sensors based on conductivity compensation of the relative permittivity can significantly improve the accuracy of soil moisture detection and the applicability to loess soils with different conductivities.