An Effective Urban Road Identification Method

Aiming at the problems of lower precision, holes and fractures in the complex environment for multispectral image road extraction, the paper proposes an urban road recognition method based on high-level semantic knowledge. It extracts the low-level features of the image and segments different objects. Establish mapping relationship between target objects and their low-level features by considering different features or feature combination. Extract the center lines for target objects, build? the semantic relationship between potential road areas, surrounding green belts and lane line objects to recognize road urban roads. Experiment results show the method can better recognize urban roads in different environments.