Frequency domain reflectometry (FDR) is a widely used technique for soil moisture measurement, but its results are susceptible to the influence of soil texture. In this paper, the effects of different soil textures (i.e., different soil particle size percentages) on soil moisture measurement by FDR were investigated on the premise that the soil organic matter content, electrical conductivity, salinity and soil temperature were fixed. Through experimental analysis, a soil texture correction model is established, which provides a theoretical basis for improving the measurement accuracy of FDR under different soil texture conditions. The experimental results showed that different soil textures had significant effects on FDR measurement of soil moisture. The FDR measurement errors of sandy soil, loamy soil and clay soil were 5.20%−7.69%, 3.45%−5.19% and 7.22%−9.04%, respectively, and the errors were significantly correlated with the soil particle size percentage. The relative error between the corrected FDR measurements and the drying method measurements was reduced by more than 2%, and the measurement accuracy was significantly improved.