Fractional Vegetation Cover (FVC) is an important parameter for the depiction of surface vegetation coverage and the monitoring of ecological conditions. Remote sensing has become the primary technique for regional FVC estimation. However, the accuracy of FVC estimation is greatly limited due to the variations in solar radiation caused by terrain and the strong heterogeneity and complexity within remote sensing pixels. This study utilized the random forest regression model to develop a fine-scale FVC estimation method for complex terrain areas by integrating surface reflectance, terrain features and view geometry (SRTVG). The Qilian Mountains region, Yellow River Source region, and Hengduan Mountains region on the Qinghai-Tibet Plateau were selected as the test areas. SRTVG was applied using Sentinel-2 imagery, and evaluated using FVC data obtained from UAV images and existing FVC remote sensing products. After incorporating terrain features and view geometry information, the new method achieved an R2 of 0.89 and a Root Mean Square Error (RMSE) of 0.13. Compared to methods not incorporating terrain features and view geometry information, the new method has reduced the RMSE of FVC estimation by 19.26%—28.02%. Compared to the existing MultiVI FVC and GEOV3 FVC products, the new method has reduced the RMSE by 40.91% and 16.67%, respectively. The newly developed SRTVG method can improve the accuracy of FVC remote sensing estimation in complex terrain areas, and enrich the technical methods for remote sensing estimation of FVC.