Linear Feature Extraction from SAR Images based on the modified LSD Algorithm

摘要

This paper proposes a new approach for linear feature from SAR images based on the modified LSD algorithm. This approach consists of three steps: computing the information of the edge, extracting the lines and the validating the lines. First, the ROEWA detector and the Gabor filter are used to extract the information of the edge, which solves the problem that the LSD algorithm cannot extract the edge and calculate the orientations accurately in SAR images. Second, the points that have the similar orientations can be grouped by region growing to acquire the line-support regions. Then we use rectangles to fit the parameters of the related lines. Finally, using the Helmholtz principle to validate the rectangles and exclude the false lines. The experiment results show that the modified approach can accurately extract linear feature from SAR images. Compared with the classical method of phase group, the proposed method can compute faster and acquire more reliable performance.