【Objective】 Soil salinization seriously restricts the sustainable development of agriculture, and the accurate monitoring of soil salinity is crucial for agricultural management and ecological protection. This study combined unmanned aerial vehicle (UAV) imaging spectroscopy with machine learning algorithms to explore the inversion and spatial mapping of soil salt content (SSC) in coastal areas. 【Method】 Feature bands were selected using the Competitive Adaptive Reweighted Sampling (CARS) algorithm, and spectral indices were calculated. Spectral indices were selected using the Recursive Feature Elimination (RFE) method. Utilizing PLSR, SVR, and RFR, this study developed prediction models for all spectral bands based on six different spectral transformations, and spectral index prediction models were built using SVR, RFR, XGBoost, and BPNN. The best model was chosen for SSC spatial mapping through accuracy evaluation. 【Result】 The results showed that the measured soil salt content (SSC) in the study area ranged from 1.23 to 8.96 g kg^-1, with a mean of 3.12 g kg^-1. Among the full-spectrum models, the random forest regression (RFR) model based on raw spectra processed with Savitzky-Golay (SG) smoothing demonstrated the highest accuracy. For the spectral index models, the extreme gradient boosting (XGBoost) model with feature selection performed the best. The inversion results revealed that low-to-moderate soil salinity was widely distributed across the study area, with high salinity values scattered sporadically. While XGBoost was well-suited for predicting the overall spatial distribution of soil salinity, the RFR model based on SG-smoothed raw spectra was more effective for mapping areas with low salinity. 【Conclusion】 This study innovatively combined full-spectrum optimized spectral indices with traditional ones to build a SSC prediction model, offering a new technical path for rapid SSC monitoring in coastal regions using UAV imaging spectroscopy.