Soil moisture, dielectric constant, and compaction are key parameters that characterize essential physical properties of soil. They play a crucial role in digital soil mapping (DSM), precision agriculture, and ecological research. This study focuses on a typical red clay farmland area in Kunming, where high-resolution multispectral UAV imagery was acquired alongside in-situ measurements of surface soil volumetric water content, dielectric constant, and compaction. Vegetation indices were calculated from the imagery, and key spectral features were selected using Pearson correlation analysis (PCA) and the Variable Importance in Projection (VIP) algorithm. These features were then used to develop quantitative inversion models of soil parameters using Multiple Linear Regression (MLR) and Random Forest (RF), followed by spatial mapping.The results show that the VIP algorithm effectively identifies spectral features with significant predictive contributions, and the RF model outperforms the MLR model in terms of prediction accuracy (R2 > 0.936). Spatial distribution maps reveal clear patterns of variation for soil moisture, dielectric constant, and compaction. Notably, moisture and dielectric constant exhibit highly consistent spatial distributions, reflecting their physical interdependence, while compaction displays more complex spatial heterogeneity. This study demonstrates that UAV-based multispectral inversion models can achieve high-accuracy synergistic estimation of red soil physical parameters, offering valuable support for soil resource digital management and sustainable agriculture.