More and more attention has been paid to extraction of roads and other linear features from high-resolution synthetic aperture radar(SAR) images. Due to the complicated background of objectives and the speckle noise in the high-resolution SAR images

it is almost impossible to extract roads directly from original remote sensing images. In order to extract roads precisely from high-resolution SAR images with complicated background and speckle noise

a method of using genetic algorithm(GA) is developed to extract main roads in the paper. After the original SAR images are filtered to suppress the speckle noise

Fuzzy C means is used to classify the images unsupervisedly into vegetation

built areas

roads and other class

and the pixels belonging to the main roads are isolated from the images to simplify the original problem. Then

according to the membership of pixels to main roads and the uniformity of gray

a road model is constructed

and the genetic algorithm is used to search globally optimized roads. The experimental results show that the presented approach can effectively extract main roads from high-resolution SAR images.