Urban road quality design supported by multi-source data and deep learning

Digital road information is not only an important basic geographic element, but also a key component of basic geographic information in China. Accurate access to digital road information is very important for traffic management, urban planning, road control, GPS navigation and map updating. At present, there are two main ways to obtain digital road information: GPS trajectory data and remote sensing image. The road target in remote sensing image has remarkable characteristics in space, radiation, topology and texture. With the popularization of deep learning, the method of road extraction from remote sensing image using deep learning has become the main research direction of road extraction. this paper supports multi-source data.