Urban road extraction from optical remotely sensed imagery

With the development of the city, the road has changed frequently. Road extraction from optical remotely sensed images is an economic and efficient way to obtain and update road networks. This paper presents a method to extract the urban road from optical remotely sensed images. Our proposed method includes the following three steps. First, an unsupervised K-means Clustering classification is applied to classify the images into two categories: road and non-road. Followed by road's homogeneous property to remove noise and improve road extraction accuracy. Finally, the shape features of road is adopted to remove other manmade objects, including buildings, parking lots and other objects which has similar spectral feature to roads. The main road networks are generated after these steps. One application of this algorithm is to construct the road network from satellite images and compare to old city map, to detect the change by urbanization.