Aiming at improving the accuracy of salient object segmentation, a salient object detection method based on edge and multi-scale feature fusion was proposed.Firstly, the ResNet50 network was used to extract features from images. Then, an improved spatial attention module was utilized to enhance the representation ability of features.Next, a novel edge and multi-scale feature fusion module was proposed, which organically combines edge information with multi-scale features information, and a loss function that comprehensively considers salient object subject segmentation and edge segmentation was designed to effectively supervise the future fusion module, ensuring that the model will simultaneously focus on the detail information of salient object subject and edge during training, in order to improve the clarity of salient object subject and edge.Finally, a context enhancement module was introduced innovatively to effectively reduce information loss during multiple upsampling and downsampling processes in deep learning networks, thereby improving the accuracy of the salient object subject and edge.Compared with eight mainstream algorithms in recent years on three public datasets, this method outperforms other algorithms in both quantitative and qualitative results, verifying its effectiveness and superiority.