Agricultural Field Border Line Retrieval Using Optical and SAR Imagery

This work introduces a workflow to retrieve agricultural field border lines from remote sensing images in a fully automatic manner. The methodology builds upon the joint usage of optical and SAR images and is composed of three main processing blocks: (1) Pre-processing of image data including a geometric adjustment of the optical images w.r.t. the highly accurate SAR data, (2) Field border line extraction based on a novel definition of stable features and a custom-tailored segmentation algorithm. (3) A feature-based matching principle that aligns the field border lines to a reference GIS dataset. The proposed framework is evaluated on TerraSAR-X (SAR) and optical RapidEye imagery, yielding sufficiently accurate agricultural field border lines.