Abstract: This paper proposes a new method that integrating texture and geometrical characteristics for the road extraction of high resolution remote sensing imagery. Firstly, texture characteristics of the high resolution remote sensing imagery is analyzed, and a clustering method based on texture characteristics is proposed. The clustering method approximately divides the imagery into road areas and nonroad areas. Then the geometrical characteristics are utilized to reject the false road pixels which have been mingled in road areas, the preliminary road structure has been formed. Finally with the function of mathematical morphology, the holes that come from vehicles and trees are wiped off. The integrated road information is obtained. Compared with the other traditional methods, the result of the experiment shows that the proposed method can extract the road information in the high resolution remote sensing imagery effectively.


Key words: road, texture, clustering, geometrical characteristic, mathematic morphology