A New Approach on Topographic Feature Point Extraction of SAR Imagery

The feature points extraction is a foundation for feature based SAR images registration, the quality of feature points directly determines the accuracy of registration. Current SAR feature points extraction algorithms are mainly shifted from optical image corresponding algorithms, accordingly feature points extracted with these algorithms are not only impacted by the speckle noise, but also not able to describe topography features effectively, especially in the mountain regions, both disadvantages of these algorithms degrade seriously the accuracy of SAR image registration. In this paper, we present a new approach to extract those special feature points in the SAR imagery, the topographic feature points, with sub-pixel level through a novel topographic feature point detection algorithm which bases on the SAR imaging characteristics of the terrain. After the analysis of the geometric and scattering characteristics of topographic features in the SAR imagery, a topographic feature point detection algorithm is designed and implemented. During the procession, the SAR image is firstly filtered by a speckle suppression algorithm, then an improved Harris algorithm is used to extract feature points. Furthermore, non-topographic feature points extracted in former steps are removed by the topographic feature points detection algorithm proposed in this paper. Finally the topographic feature points which can reflect terrain characteristic are reserved. The experiment results indicate that the approach put forward in this paper is effective in extracting topographic feature points.