Abstract
To solve the problem that the road information in high-resolution remote sensing image is confused with the surrounding environment and interfered by non road information such as trees, buildings and shadows, this paper proposes an improved road extraction model by adding Non-local block in D-Linknet encoder to capture remote dependency, which can improve the shadow-shielding-road recognition ability. The experimental results show that this method can effectively reduce the interference of non road factors and significantly improve the effect of road extraction. This method decreases the fragmentation of road patches during extracting, and keeps the integrity of the roads,so it can be used for smart manufacturing and mapping in high-resolution remote sensing image.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
