Road Extraction from High-Resolution SAR Image on Urban Area

Abstract

Because of active and side-look imaging and speckles in the SAR image, the road edge is blur and it is difficult to determinate the road edge. Therefore, extraction of the road from high-resolution SAR image can't use the methods that were used to the optical image. In this paper, we research how to extract the information of zonal road and how to get the vector road from the high-resolution SAR image. Usually, urban road is regular. In remote sensing image, the road has many characters, such as functional character, spectral character, geometric character and texture character so on. These characters consist of the knowledge of road extraction. After accurately expressing the character knowledge with mathematical equations, we can extract the road information more accurately. Therefore, we must construct road model after analyzing the road characters in the high-resolution SAR image. Based on the above analysis, we may accurately extract road from SAR image. By our experiment, the result proves that our method is a good idea.