Salient Object Detection in Optical Remote Sensing Images (ORSI-SOD) aims to segment ground objects with salient visual features from optical remote sensing images with complex backgrounds. Although various methods have been proposed in the field of ORSI-SOD, several challenges remain, such as insufficient effectiveness in feature extraction, irregular distribution of targets, and local occlusion interference caused by complex background disturbances. To tackle these obstacles, we put a Salient Object Detection in Optical Remote Sensing Images Using the Segment Anything Model with Feature Augmentation Network (SF-Net). First, we utilize the pre-trained Hiera backbone of the Segment Anything Model 2 (SAM2) for feature extraction in our network. Next, we introduce the Multi-scale Attention Enhancement Module (MAEM), which enhances feature representation by extracting features at multiple scales and optimizing attention mechanisms, allowing salient objects to be effectively distinguished from the background. Subsequently, we propose the Detail Supplementary Fusion Module (DSFM) for SF-Net, which boosts the model’s competence in perceiving detailed features in complex scenes by hierarchically integrating edge-aware representation learning and cross-scale feature fusion mechanism. Finally, we administer extensive experiments on three benchmark datasets (EORSSD, ORSSD, ORSI-4199). The analysis findings prove that SF-Net outperforms existing methods, and we analyze the potency of the submitted subsystems through ablation studies.