Abstract: It is extremely difficult to extract road from high-resolution SAR images because of the existence of various kinds of disturbances, occlusions and shadowing. A new road center-point extraction method is proposed by combined using local detection and global tracking. In local detection phase, two windows are set. The outside window is used to obtain the local road direction by using non-linear structure tensor, based on the fact that fences, green belts and other disturbances point to a consistent direction with the edge of roads. The inside window adjusts its direction by the result of non-linear structure tensor. Then it searches for the road areas, and determines the width and center of roads. In global tracking phase, particle filter of variable-step is used for solving the problem of tracking broken frequently by occlusions on the road and shadowing alongside the road. In 1 m high-resolution airborne SAR image experiment, the results indicate that this method is effective.