Linear features’ detection in SAR images using Fuzzy Edge Detector

Abstract

Synthetic Aperture Radar (SAR) images exploitation is of great importance in remote sensing and Earth observation techniques, as well as in military surveillance and reconnaissance missions. One of the most important image-processing techniques is image segmentation. Segmentation subdivides an image into its constituent parts or objects. Edge detection is often an important first stage in many types of image segmentation. It can be used to simplify complex imagery in preparation for subsequent feature identification. Detection of linear structures (road network, airfields e.t.c.) is of great interest in a potential automated recognition system. A new approach for edge detection in SAR images, which overcomes the necessity of choosing arbitrary thresholds is described in this paper. The Fuzzy Edge Detector (FED) not only discriminates between edge pixels and no-edge pixels, but assigns each image point a degree of being an edge point. This intermediate fuzzy output enables a very flexible post-processing and thus an efficient edge detection processing chain (e.g. non-maximum suppression and edge linking). Additionally, the edge direction information resulting from FED can be used for linear features’ identification and extraction. The performance of the proposed edge detector on airborne and spaceborne SAR images that include linear features is presented.