Road Centerlines Extraction from High Resolution Remote Sensing Image

The acquisition of road network information based on high-resolution remote sensing images has important practical application value. This paper focus on the phenomenon of complex background influence and easily produce spurs around the true road centerlines, the method of using convolutional neural network for road region extraction and segmental smoothing road centerline is proposed to overcome the above problems. The experiment show that the method of this paper can overcome the influence of complex background of high-resolution remote sensing images and accurately extraction the road centerline that conforms to the actual road shape, which has great practical application value.