Road extraction from high-resolution remote sensing imagery based on local adaptive directional template match

An approach of road information extraction for the high-resolution remote sensing imagery was proposed based on local adaptive directional template match. The remote sensing imagery was segmented by means of the local adaptive directional template match and a binary image of the coarse road was obtained. Guided by the principle of local gray-scale consistency, the consistency of the input images was tested and the test results on local gray-scale of all the pixels in the image were received. Afterwards, the preliminary segmented results and the results of local gray-scale consistency test were fused to get a better original part of the road. Finally, the results after information fusion were filtered according to the different features of the road shapes and the original part of the road was attained. The areas of the road in the image are not too small as the road is successive from the space aspect. Accordingly, the parts of smaller area were removed on the basis of area index and non-road information was also removed based on the length-width ratio, so that the exacted results were obtained. The algorithm was verified by two high-resolution remote sensing images and it was compared with known road information exaction methods were accomplished. The results show that the road information can be more successfully exacted by this approach as shown in the experiment results.