Abstract: In high-altitude underground mining operations, the soil moisture content at mine outlets is unevenly distributed and easily affected by rainfall seepage. The moisture variation from shallow to deep layers is a key factor inducing debris flow or mud inrush disasters. Traditional contact measurement methods cannot effectively monitor this area. To address this problem, a soil moisture inversion method based on Stepped Frequency Continuous Wave (SFCW) ground-penetrating radar and Support Vector Machine (SVM) was proposed to measure the soil moisture content at mine outlets. The method set the detection depth at 0.3 m, and achieved rapid and non-contact measurement of soil moisture content at mine outlets through radar scanning, signal preprocessing, feature extraction (reflection coefficient and phase difference), and SVM modeling and inversion. Simulation results showed that SFCW radar effectively identified reflection features caused by moist regions, verifying the feasibility of this method in detecting layered and heterogeneous soils. Under laboratory conditions, SFCW radar was used to measure sandy loam samples collected from the Pulang Copper Mine outlet, and by comparing multiple modeling methods, the SVM model based on the combined features of the reflection coefficient and phase difference was found to produce the best inversion performance. Field measurements at the Pulang Copper Mine outlet showed that the relative error between predicted and measured values was within the range of 9.67% to 14.53%, indicating that the method provides relatively reliable soil moisture estimation within an effective detection range of 1 m horizontally and 0.3 m in depth.