Soil moisture content (SMC) plays a vital role in seed germination and crop growth. It is of great significance for precision agriculture to acquire the SMC of seed-dropping point in planting for the sake of decision-making and depth-regulating of seeding. Thus, developing a proper SMC sensor will contribute a lot to precision agriculture. An SMC sensor was designed by using visible and near-infrared (VIS-NIR) light source. The spectral data of soil samples was collected by a high-resolution spectrometer, then the partial least squares regression (PLSR) was used for determining the optimal pretreatment method, and various dimensionality reduction methods were employed to select the characteristic wavelengths of soil moisture. It was concluded that the sensitive reflectance bands of different SMC within 400~1000nm were around 410nm, 540nm, 780nm and 970nm. Through the modeling analysis of combinations of two of these four wavelengths, the optimal wavelengths of VIS-NIR light sources for prediction were selected as 410nm and 970nm, respectively. The results of experiments conducted in the laboratory showed that when the distance between the sensor and the measured soil surface was under 3mm, within the range of 0.69%~28.45% SMC, the predicted and the measured values appeared a justified linear correlation for which the coefficient of determination (R 2 ) was 0.81 while the root mean square error (RMSE) was 2.90%; within the range of 0.69%~22% SMC, the R 2 of the linear model reached 0.93 and the RMSE was decreased to 1.72%. The factorial test indicated that temperature and light scarcely had influence on the SMC sensor at 0.05 level. The results of simulated field tests indicated that rocks and the process of acquire soil sampling may generate outliers. The R 2 of the linear correlation reached 0.82 and the RMSE was 1.23% after the outliers were excluded, which met the requirement of SMC detection in most conditions of precision agriculture such as maize planting.