Research And Improvement Of Urban Road Extraction Algorithm In High Resolution SAR Images

Traffic is not only a bond of the city but also an important channel for the development of the city transportation of people and goods;urban road is important for the construction of urban traffic system,so,it is essential to obtain urban road information quickly and accurately.Because SAR(Synthetic Aperture Radar)has the advantages of all-day and all-weather,it is of great practical value to extract urban road information from SAR images.High resolution SAR images are able to display the urban road characteristics.However,since the high resolution SAR urban road images are disturbed by the buildings,trees and other information,coupled with the impact of the inherent noise of SAR images,the road images are so prone to fracture that the extraction process becomes very complex.Therefore,SAR image denoising algorithm,road extraction algorithm and road global connection algorithm are studied to get and update the information of urban road promptly so that realizing information management of urban traffic.The main work is detailed as follows:Firstly,in order to reduce the influence of speckle on the road information of SAR images,a local adaptive Bayesian threshold filtering algorithm is proposed based on the Contourlet transform after the research of the Contourlet threshold filtering algorithm.The tested result shows that the proposed algorithm has a better filtering effect and can preserve more detail information of SAR images than the traditional speckle reduction algorithm.Secondly,combining the advantages of multi-feature fusion algorithm and fuzzy connectedness algorithm,an improved multi-feature fusion algorithm is described.Compared with the traditional multi-feature fusion road extraction algorithm,The tested result shows that the proposed algorithm can improve the integrity and accuracy of road extraction.Thirdly,a new line element acquisition algorithm is proposed,which can produce the road line element information for the input of global connection algorithm to get road network and improve the integrity of road extraction.Finally,a prototype system of urban road extraction for high resolution SAR images is designed and implemented by integrating the related algorithms mentioned above.After testing,the system works properly.