Local Road Area Extraction in CSAR Imagery Exploiting Improved Curvilinear Structure Detector

Abstract

Road extraction is an important part of synthetic aperture radar (SAR) image interpretation. In recent years, circular SAR (CSAR) has attracted extensive attention from researchers owing to its ability of 360-degree observation. Due to the unique imaging geometry of CSAR, CSAR images contain more complete road information. However, the curvilinear-structured appearance of roads in CSAR images and the complexity of the scene result in difficulties in the road extraction. The Curvilinear structure detector (CSD) is capable of extracting the curvilinear structures with a specific width from complicated image scene. Based on the traditional CSD, an improved CSD (ICSD) for local road area extraction from CSAR images is introduced in this paper. First, by adopting ICSD, the edges of a CSAR image and centerlines of local roads are extracted, as well as their direction. Second, the local roads are obtained by geometrical and radiometrical rules. Finally, the missing intersections and edge pixels are repaired to acquire high-precision and high-quality extraction results of local road area. The experimental results on different band CSAR images reveal that the proposed method exhibit enhanced performance than the three state-of-the-art methods.