Road Extraction from Satellite Images Using Maximum Entropy Threshold Method

Traffic management is important for sustainable urban development, smart cities, traffic navigation, and urban planning. With the development of modern aerospace technology, remote sensing technology, and computer technology, the use of computer vision to extract roads from satellite remote sensing images has become the main means of road information acquisition. Nowadays, satellite road extraction is widely used for maps and smart cities, so it is very necessary to carry out accurate road extraction. Satellite images often contain noise and multiple objects, which can affect the edges of the extracted roads. This paper proposes a method that is based on considering the geometric characteristics of the road. It uses the characteristics of small changes in the geometric characteristics of the road and obvious edge information on both sides of the road. The road surface graphics with regular shape characteristics are obtained by the maximum entropy threshold method, dilation processing, skeleton processing, burring, and erosion. Experimental results show that road information can be extracted accurately with high precision.