In the remote sensing images, roads are considered to be the long and narrow linear target which is similar in color, texture and shape. Based on these features, the entire road network in the image will show a very significant feature, which can easily excite the attention of the human, which can be called the region of interest. The region of interest(ROI) in the scene can cause the most interesting of users, which reflects the main content of the image area, visual cognitive theory study shows that the visual attention mechanism can simulate the observation processing of the human eye to identify the salient region of remote sensing images. This paper proposes the idea of using visual attention mechanisms to assist road network extraction by analysising and processing the salient region and get the final road network. Comparative experiments show that the algorithm can effectively improve the accuracy and integrity of the road network extraction.