Abstract: Soil salinization can seriously impact the agricultural productivity in the saline-alkali lands. It is vital to timely and accurate monitor the soil salinization dynamics for the amelioration of soil degradation. Fortunately, the remote sensing can be expected to monitor the regional soil salinity in recent years. The salinity/vegetation remote sensing indices have also been widely used for the inversion of soil salinity. However, the accuracy of these indices is often compromised by the great variations in the geographic and crop cover environments, leading to the substantial uncertainty of the inversion. This study aims to enhance the precision of regional soil salinity inversion using remote sensing. Multi-temporal Sentinel-1/2 sensing data was integrated with the remote sensing indices under different crop covers. The study area was set in the Yellow River Delta region in Shandong Province, China. The concurrent ground sampling was employed to analyze the sensitivity of existing soil salinity inversion indices. The new indices of remote sensing were then constructed to specifically design for the different types of crop cover. A correlation analysis was performed on the existing soil salinity data and 35 remote sensing spectral indices. Furthermore, a series of operations were performed to filter the salinity big data in the remote sensing combinations of bands 2, 3 and 4. The results show that: (1) Four remote sensing indices were identified with the higher correlations with the soil salinity. These included the “Blue-Near Infrared” salinity index (Salinity index 2, S2), the Normalized Differential Salt Index (NDSI), the Canopy Response Salinity Index (CRSI), and the Enhanced Normalized Differential Vegetation Index (ENDVI). (2) Various combinations of 2, 3, and 4-band indices were screened for the soil salinity data. The salinity indices were obtained for the crop cover (SIYRDveg) and bare soil (SIYRDbare). The highest correlation of 0.77 was achieved over the multiple temporal phase. There was the strong correlation between remote sensing indices and soil salinity. (3) The salinity indices were combined with the Random Forest. The salinity inversion accuracy for the crop-covered surfaces was achieved an RMSE of 2.23 g/kg and an MAE of 1.25 g/kg, while the RMSE and MAE were 1.00 g/kg and 0.73 g/kg, respectively, for the bare soil surfaces. The accuracy of multiple bands was improved by 31%, compared with the only basic bands of remote sensing. Remote sensing indices and machine learning were integrated to enhance the accuracy of soil salinity mapping. Thereby more reliable data was provided for the agricultural management and environmental monitoring. The inversion of salinity indices was improved in regional cultivated lands, particularly in the areas with the diverse vegetation covers. The Sentinel-1/2 data with these newly-developed indices were significantly enhanced the precision of the soil salinity monitoring. The remote sensing can also offer a more reliable approach to manage the soil salinity in agricultural landscapes. Bare soil surface shared an increasing trend in the soil salinity, due to the evaporation and salt migration. While the vegetated surface can reduce the evaporation to enhance the moisture retention, where the salt accumulation can be suppressed to increase the salt absorption. Remote sensing indices can also be expected to tailor for the complex environmental challenges in smart agriculture.